The food system is a primary sector of the economy that accounts for 34% of total anthropogenic GHG emissions (1). Hence, reducing the related GHG emissions is at the heart of the European Green Deal (2) that, through the Farm to Fork Strategy (3), aims to make food systems fair, healthy, and environmentally friendly. In more detail, one of the goals is to ensure food security, nutrition, and public health, making sure that everyone has access to sufficient, safe, nutritious, and sustainable food. This corresponds to the promotion of sustainable diets as defined by the Food and Agriculture Organization of the United Nations (4), that is “diets with low environmental impacts which contribute to food and nutrition security and healthy life for present and future generations. Sustainable diets are protective and respectful of biodiversity and ecosystems, culturally acceptable, accessible, economically fair and affordable; nutritionally adequate, safe and healthy; while optimizing natural and human resources.”

Adopting a sustainable diet can effectively help to reach the goal since diets and consumer behaviors strongly affect the food system. In fact, because of the relationship between supply and demand, the choices of consumers drive what, and how much, is produced. For example, the choice of locally sourced and organic food helps reduce food miles and the use of pesticides, fertilizers, and antibiotics. In general, consistent evidence indicates that a dietary pattern higher in plant-based foods (e.g., vegetables, fruits, legumes, seeds, nuts, whole grains) and lower in animal-based foods (especially red meat), as well as lower in total energy, is both healthier and associated with a lesser impact on the environment (5–7). For example, the replacement of a meat meal with a vegan or a vegetarian meal significantly helps to lower GHG emissions (8).

Diets and consumer behaviors, that is food preference and preparation techniques, are greatly affected by traditions, beliefs, and values shared by a community. They define the structure of each meal and the set of foods and dishes that are appreciated and considered edible and acceptable (9). Cultural habits are quite rooted in the population and therefore a shift in consumer behaviors may require some effort. In this respect, school is a privileged environment for acquiring and solidifying eating habits through health and nutrition education as well as environmental awareness.

School lunch programs have contributed significantly to shaping dietary habits (10, 11) in line with dietary recommendations. Indeed, school meal menus are classically designed to promote healthy eating habits aimed to prevent overweight and obesity and their consequences in terms of the incidence of non-communicable diseases (12). Italian National guidelines for school feeding programs are provided by the Ministry of Health (13). These guidelines address nutritional issues and also introduce special therapeutic diets, for example for food intolerance and allergies. Moreover, intercultural menus are promoted to accommodate the eating habits of an increasing foreign school population, and special menus are guaranteed for ethical, religious, and cultural reasons. Therefore, vegan and vegetarian menus, as well as menus with religious restrictions such as pork-free recipes, can be requested. Further indications about the acceptability of the menus, and consequently the reduction of food waste, were released in 2018 (14).

The possibility of choosing either vegan or vegetarian menus, as well as the concern about food waste reduction, are the only items in the guidelines that, though indirectly, pursue the goal of reducing GHG emissions. This goal can be indeed further pursued by taking explicitly into account the GHG emissions in the design of school menus.

To this end, a suitable design tool has been recently developed (a detailed description can be found in (15)). The tool makes use of a binary (0–1 integer) programming method to provide the sequence and composition of daily lunches that, over a given period, minimize the GHG emissions needed for their preparation. The meals composing the lunches are chosen using dishes out of a given recipe book. Therefore, instead of defining just a level of consumption of food groups or food items, the tool provides a realistic menu. To cope with the nutritional recommendations, each dish is characterized by its energy and nutritional content. Furthermore, the variety of the menu can be explicitly taken into account in a very natural way, that is bounding the daily, weekly, or total repetitions of single dishes and dishes in the same food group. This is a basic issue to make the menus acceptable. Acceptability is however a broader concept implying the food preferences of the pupils. This is implicitly included in the optimization model since only the actual recipes from school menus are considered.

The tool has been used to design school lunch menus with minimal carbon or water footprint. In more detail, in (16) such a design was performed using the recipe book of the Municipality of Rome, while a sample of 52 Italian school menus from Italian macro-regions (North, South, Center, and Islands) was used in (17). The results show that a significant reduction of the environmental impact can be obtained while satisfying all the nutritional and acceptability recommendations.

This study aims to use the above tool to compare the GHG emissions associated with different kinds of menus, such as vegan, lacto-ovo vegetarian, and omnivorous, for a two-week schedule for the primary school lunch. To this end, an extension of the recipe book in (17), including vegan recipes, was considered. These recipes are not included in the standard school lunch menus but are provided only by demand.

The data consist of a set of dishes, whose composition and serving sizes are fixed, for which the energy, nutrient contents, and carbon footprint are available. The schedule must be organized by choosing within the given set of dishes the sequence of daily lunches while satisfying some constraints related to adequate energy and nutrient intakes and food acceptability. These constraints are related to the composition of the lunch (first and second-course dish, side dish, bread, and fruit), daily and weekly bounds on energy and nutrient intakes, and bounds on weekly and total repetitions of dishes in the same food group.

It was not possible to design a vegan menu satisfying all the nutritional constraints of Table 2. This is due to the lack of vitamin B₁₂ content in most vegan dishes since no fortified foods are considered. This is a well-known problem in vegan diets, usually solved by either considering fortified foods or by taking supplements. The vegan menu is then designed without constraints on the vitamin B₁₂ intake and assuming the use of appropriate supplements for it and the breakfast intake of calcium. Moreover, to promote vegetable consumption, vegetable side dishes must be offered a minimum of four times per week. The optimal menu is given in Supplementary Table SM4 and has an average weekly GHGE equal to 1,742 g of CO_2eq.

The vegetarian menu satisfies all the constraints given in Table 2, including that on vitamin B₁₂ intake. In addition to the vegetable dishes frequency of the previous menu, both dairy and egg dishes must be proposed at least once a week but no more than twice a week and three times on the whole menu. The optimal menu is given in Supplementary Table SM5 and has an average weekly GHGE equal to 2,570 g of CO_2eq.

In addition to the frequencies already fixed for vegetables, eggs, and dairy dishes, the first omnivorous menu must not contain red meat dishes, while white meat dishes must be proposed at least once a week but no more than twice a week and three times in the whole menu. The optimal menu is given in Supplementary Table SM6 and has an average weekly GHGE equal to 2,678 g of CO_2eq.

The second omnivorous menu must contain, in addition to the food groups frequencies of the previous menu, a red meat dish once a week and a beef dish once in the whole menu⁶. The optimal menu is given in Supplementary Table SM7 and has an average weekly GHGE equal to 3,578 g of CO_2eq.

A comparison of the nutritional content and carbon footprint for the above optimal menus is given in Table 3 and illustrated in Figure 1. Vitamin B₁₂ has a daily lower bound of 0.35 μg and hence the value for the weekly lower bound in Table 3 is just five times the daily one and, as previously discussed, is not satisfied by the vegan menu.

As Figure 1 makes clear, the GHGE doubles from the vegan optimal menu to the omnivorous optimal menu of the second kind. This variation mostly depends on the second-course dishes, since all the menus share similar first-course and side dishes. The carbon footprint significantly increases when passing from the vegan optimal menu to the vegetarian optimal menu and from the omnivorous optimal menu of the first kind to that of the second kind. The first GHGE variation (+48%) is due to the addition of eggs and dairy dishes to the diet; the second variation (+34%) is due to the addition of red meat, in particular beef meat. Interestingly enough, the GHGE related to the vegetarian optimal menu and the omnivorous optimal menu of the first kind is quite the same, because only fish and white meat dishes were added to the menu. This is made evident by the first picture in Figure 2, where the GHGEs associated with the second-course dishes are depicted. Indeed, eggs and dairy, fish, and white meat dishes show similar GHGE values. On the other hand, vegan dishes have the lowest impact while ruminant meat dishes have the highest one.

Let us now consider the nutritional characteristics of the four optimal menus. The weekly average values of energy, carbohydrates, fat, and sugars are quite the same. They show a deviation of a maximum of 10% from their average on the four menus. The other nutrients and minerals show more evident variations. In particular, as shown in Figure 1, fibers and iron decrease when passing from one menu to the other, with a total variation equal to −15% from the vegan optimal menu to the omnivorous optimal menu of the second kind. As above, this is also made evident by the second and third pictures in Figure 2 that show how vegan second-course dishes, consisting mainly of legumes, are richer in fiber and iron. On the contrary, as shown in Figure 1, the intake of protein and sodium increases when passing from one menu to the other, with a total variation equal to +40 and +26%, respectively. Again, the protein increase can be explained considering the protein content distribution over the dishes, given in the fourth picture in Figure 2.

Note that the average values for all the menus of energy and nutrients over the 2 weeks do not attain the maximum weekly allowed values (save for Fiber in the vegan menu). This means that the optimal menus do not provide maximum values for all the weeks. Should it happen and be an issue, two model improvements are possible to deal with it: either narrow the weekly ranges or add a further range on the whole menu.

As a general remark, we can observe that the vegan optimal menu has the lowest carbon footprint and that the introduction of just eggs and dairy in the diet significantly worsens the impact. Although the adoption of a vegan diet helps in reducing the carbon footprint of a menu, it is the optimal selection of dishes that ultimately makes the difference. Indeed, the GHGE associated with a generic vegan menu can be much greater than that associated with the vegan optimal menu. The maximum GHGE of a menu which is nutritionally adequate can be computed by just maximizing the total amount of CO_2eq emission instead of minimizing it. In the case of vegan menus, this results in a menu (see Supplementary Table SM8) with an average weekly GHGE equal to 2,937 g of CO_2eq which has an impact greater than the vegetarian optimal menu and the omnivorous optimal menu of the first kind, respectively (see Supplementary Tables SM5, SM6). Hence, eating vegan is not sufficient to guarantee a pressure reduction on the environment, but it is also necessary to make an appropriate selection of dishes.

Similar arguments hold when considering vegetarian menus. In particular, the most impacting vegetarian menu (see Supplementary Table SM9) produces an average weekly GHGE equal to 4,723 g of CO_2eq which is far higher than that produced by the omnivorous optimal menu of the second kind (see Supplementary Table SM7). In other words, a non-optimal choice of dishes in the design of a vegetarian menu may lead to an environmental pressure higher than that of some omnivorous menus.

Figure 3 makes clear these results by showing that the possible values of GHGE associated with generic vegan, vegetarian, and omnivorous menus partially overlap. This means that there are vegan menus producing pressure on the environment higher than some vegetarian menus. Similarly, there are vegetarian menus that exhibit higher GHGE than some omnivorous menus.

In general, adding fish and non-ruminant meat to the vegetarian menu does not change its overall impact, but ruminant meat consumption determines a further worsening. The impact increment depends on the meat food groups introduced and on their frequencies on the menu. Nevertheless just optimizing the menu, i.e., the optimal selection of the dishes composing the menu may reduce significantly its carbon footprint. This is made evident by computing the carbon footprint of the current school lunch menu proposed by the municipality of Rome. This menu is defined over 9 weeks and has nutritional values in line with the daily and weekly bounds given in Table 2⁷. It has been developed by nutritionists and it is guaranteed to be nutritionally adequate and balanced in carbohydrates, fats, and proteins, along with all essential minerals, vitamins, and health-promoting substances. It is quite rich in every food group and can hence be compared with the omnivorous optimal menu of the second kind that contains exactly 1 second-course dish with fish, one with red meat, one with white meat, one with eggs, and one with dairy, every week. The weekly GHGE average of the menu of Rome is equal to 5,841 g of CO_2eq so the omnivorous optimal menu shows a reduction by about 40% of the carbon footprint. This is precisely in line with the target of the European Green Deal⁸ and is mainly due to the optimal selection of dishes made in the design of the menu. For example, the two red meat dishes on the menu are a dish with pork (roast pork) and a dish that contains beef together with pork (beef/pork meatballs). Indeed, the menu must contain one red meat dish once a week and one beef dish once in the whole menu, and roast pork and beef/pork meatballs are the dishes with minimal GHGE among all the red meat dishes and red meat dishes with beef, respectively.

The properties of the menus considered in this paper are similar to those at the single recipe level presented in (22), where a total of 311 recipes from six different German omnivorous, vegetarian and vegan cookery books were analyzed in terms of GHG emissions and costs. Moreover, the results are in line with the observational studies on the adult population made in France (8) (using a food frequency questionnaire over about 29,000 participants), the UK (32) (using a food frequency questionnaire over about 55,500 participants aged 20–79), and Italy (33) (using a 7-days dietary record over 153 participants), in which the environmental impacts of different diets with different levels of animal product consumption were compared. Similar results were obtained by designing diet plans using linear programming (34, 35). In particular, the diet plans are obtained by minimizing their deviation from a reference/current diet while fulfilling nutrient and climate footprint constraints. Constraining the diet GHG emissions results in a reduction in meat, dairy products, and processed foods.

In general (36), the results available in the literature show that diets and observed dietary patterns in the population are less impacting on the environment the higher they are rich in plant-based foods. In more detail, the results suggest that the vegan diet is the optimal diet for the environment because, out of all the other diets, its production results in the lowest level of GHG emissions. Nevertheless, there is the possibility of significantly reducing the environmental impact of a given diet, without excluding the meat and dairy food groups, but rather, by reducing them substantially.

Our findings confirm and enhance these results. Indeed, they show that the above properties of diets and observed dietary patterns remain true when considering school lunch menus with minimal GHGE. This conclusion is not obvious since the tool used in this work provides realistic menus in place of diet plans. Moreover, it is shown that although the adoption of a specific diet may help in reducing the carbon footprint, it is the optimal selection of dishes that ultimately reduces the impact of a menu.

This study shows that, in general, a low carbon footprint menu is plant-based. It also results that meat consumption, and in particular ruminant meat, is the most important contributor to the carbon footprint within the protein source dishes. Hence, meat dishes should be partially replaced with legumes, cereals, fruits, and vegetables. Anyway, to comply with cultural habits, it is possible to include on the menu fish and white meat dishes that are nutritionally beneficial and more sustainable than red meat dishes.

The added value of the proposed methodology, based on an integer optimization approach, is that the choice of a particular diet (vegan, vegetarian) does not, per se, guarantee a GHGE reduction as compared to a generic menu: it is the optimal selection of dishes on the menu that determines the minimal carbon footprint. Moreover, the optimal choice of dishes on omnivorous menus can lead up to a 40% emission reduction compared to the current school lunch menu of the municipality of Rome. This allows to progressively shift food behavior toward the direction of environmental protection, avoiding drastic changes to achieve environmental objectives step by step.

This result could be further improved by considering aspects related to the promotion of food short supply chain and seasonal products, as well as agricultural and processing techniques (37).

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

The research questions, conceptualization, design of the study were carried out, data were analyzed, software was designed, and writing and original draft preparation was carried out by LB and AD. The methodology was revised and database compilation was carried out by MF, DM, and LR. Data interpretation was undertaken and writing, review, and editing were done by LR, MF, DM, LB, and AD. All authors have read and agreed to the published version of the manuscript.

This paper was funded by SYSTEMIC An integrated approach to the challenge of sustainable food systems: adaptive and mitigatory strategies to address climate change and malnutrition, within the Knowledge Hub on Nutrition and Food Security, which has received funding from national research parties in Belgium (FWO), France (INRA), Germany (BLE), Italy (MIPAAF), Latvia (IZM), Norway (RCN), Portugal (FCT), and Spain (AEI) in a joint action of JPI HDHL, JPI-OCEANS, and FACCE-JPI launched in 2019 under the ERA-NET ERA-HDHL (No. 696295).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.