Trends in food consumption in Italy and analysis of WF

For the analysis of food consumption trends in Italy we used the data of the periodical household surveys provided by the Italian Institute of Statistics (ISTAT). We analyzed data of available years over a time span of 16-years, specifically: 2005; 2009; 2013; 2017; 2021. Data provided by the survey were representative of the Italian population, with a sample ranging between N = 45,000 and N = 50,000 (with the only exception of year 2013, when the sample was N = 20,275).

The ISTAT survey gathered information on food consumption by utilizing various questions about how often individuals consumed 14 primary food categories:

• • bread, pasta and rice;

• • potatoes;

• • fruits;

• • leaf vegetables;

• • vegetables;

• • red meat;

• • white meat;

• • fish;

• • pulses;

• • milk;

• • cheese;

• • cured meat;

• • sweets;

• • snacks.

The survey did not include data on individual consumption of beverages, olive oil, and eggs. As such, the estimated average calorie intake of RID is approximately 1600 kcal day⁻¹, which is a bit lower compared to the recommended daily calorie intake of the IMD (around 2000 kcal day⁻¹) For the sake of comparison, we kept the calorie intake constant at 1600 kcal day⁻¹ in all diet scenarios.

Participants in the survey were requested to disclose their eating habits in terms of how often they consumed various foods, using a semantic scale ranging from ‘more than once a day’ (= 1) to ‘never’ (= 5). The consumption frequencies were then switched into weekly portion sizes, measured in grams that subsequently let us calculate the individual weekly diet.

After analyzing variations in food consumption of the specific food categories across the considered years, we focused on the most recent consumption data (2021) and estimated the WF of the RID. The latter was calculated by multiplying the unit water footprint (uWF, i.e., the quantity of water needed to produce a unit amount of the product) by the daily individual consumption (in grams) of that food. The same approach was adopted to calculate the WF of the IMD. The WF of the RID was then compared to the one of the IMD to evaluate differences and the factors influencing them.

The WF data—the CWASI database and the WF of fish

The first and most widely used WF database, named WaterStat, has been created by the Water Footprint Network (Mekonnen and Hoekstra, Reference Mekonnen and Hoekstra2010a, Reference Mekonnen and Hoekstra2010b). The WaterStat database encompasses average uWF measurements for both the green and blue water components spanning from 1996 to 2005. These measurements are available for a variety of agrifood products, originating from both crops and animals, including both primary and processed goods. The uWF, expressed in m³ t⁻¹ or, equivalently, in L kg⁻¹, quantifies the volume of water needed to produce a specific quantity of products.

Tamea et al. (Reference Tamea, Tuninetti, Soligno and Laio2021b) improved this database within the CWASI—Coping with water scarcity in a globalized world—project (project funded by the European Research Council (ERC-2014-CoG, project 647473) and led by Prof. F. Laio from Politecnico of Turin, Italy), the one from which we retrieved the WF data for the present study. The novelty of the CWASI database is represented by a differentiation between the production and supply side of annual values of uWF of primary and processed crops.

The uWF of production (uWFp) pertains to crop products that are cultivated locally, and it refers to factors such as evapotranspiration and crop yield. These values are estimated on an annual basis, beginning from 1961 to 2016, and are available for a total of N = 255 countries.

The uWF of supply (uWFs) represents the domestic provision of primary and processed crops, originating from both local production and international trade. These figures are calculated as an average value between the quantities of local production and imports from the year 1986 to 2016 (see Tamea et al., Reference Tamea, Tuninetti, Soligno and Laio2021b for a comprehensive review).

The CWASI database kept the WaterStat values for uWF of animal-based foods, with no temporal variation.

For the purpose of our analysis, we used data of uWF of supply both for crop and animal-based products. This choice has been made since the supply data better represent human consumption, which is the focus of this research. We used data of uWFs of crops of the year 2016, the most recent available.

For the assessment of the WF of fish we used the study from Pahlow et al. (Reference Pahlow, van Oel, Mekonnen and Hoekstra2015), like previous studies (see for instance Vanham et al., Reference Vanham, Guenther, Ros-Baró and Bach-Faig2021), since the WF of fish is not included either in the WaterStat database or in the CWASI one. The motivation is that computing the WF of marine fish like any other agrifood products, meaning calculating it as proportion between the average worldwide ocean evaporation and the total amount of caught fish (Fereres et al., Reference Fereres, Villalobos, Orgaz, Minguez, van Halsema and Perry2017), would return an unreasonably high value, besides being devoid of significance, given the fact that ocean evaporation (a component of the calculation) is a process that would occur anyway (Fereres et al., Reference Fereres, Villalobos, Orgaz, Minguez, van Halsema and Perry2017). Nonetheless, it is possible to approximate the WF of aquaculture fish by assessing the amount of water used and polluted during the production of their food. This methodology mirrors the approach used for determining the WF of various other animal products.

We considered the blue and green WF values of aquaculture fish referred to the year 2008 from Pahlow et al. (Reference Pahlow, van Oel, Mekonnen and Hoekstra2015).

The Sustainable Diet Model and alternative dietary scenarios

To examine different dietary options that can reduce the WF while maintaining nutritional adequacy, we employed a modified version of the sustainable diet model (SDM), originally developed by Cavaliere et al. (Reference Cavaliere, De Marchi, Frola, Benfenati, Aletti, Bacenetti and Banterle2023).

By adapting the formula, the SDM solves the following problem:

(1)$$\arg \min _{\matrix{ {m_i \le x_i \le M_i, } \cr {\sum\nolimits_{i = 1}^{14} {k_ix_i = K} } \cr } }{\rm \omega }_w\mathop \sum \limits_{i = 1}^{14} \displaystyle{{w_ix_i} \over {1\left[{\displaystyle{{{\rm liter}} \over {{\rm pers}\cdot {\rm day}}}} \right]}} + {\rm \beta }\mathop \sum \limits_{i = 1}^{14} \left({\displaystyle{{x_i-p_i} \over {\,p_i}}} \right)^2$$

In equation (1), the vector x _i = (x ₁, x ₂, … , x ₁₄) is the weekly consumption, measured in grams, of all 14 food categories included in the analysis. The values m_i and M_i represent the lower and upper bounds for the range of intake for each food category. The variable K is the total weekly caloric intake of the diet and it is constant. The variable k_i represents calories of each specific food category and it can vary depending on the chosen diet model being tested.

The variable w_i is the WF associated with the weekly consumption of each food category. The first term, $\sum\nolimits_{i = 1}^{14} {w_ix_i}$ equals the total WF of the diet.

The term ${\rm \beta }\sum\nolimits_{i = 1}^{14} {{( {( {x_i \hbox{-} p_i} ) /p_i} ) }^2}$ quantifies people's resistance to modify their dietary patterns when substantial variations are involved, such as the complete removal of specific food categories from the diet. Higher values for β indicate that people have a low acceptability of new dietary patterns, while low β means a greater willingness to alter the diets.

Therefore, the solution to equation (1) seeks to find best dietary solution that minimizes both the total WF and the deviation from the RID. In this equation, the parameter β is set to different values: 0.2 for the mainly animal-based diet, 0.6 for the mainly plant-based diet, and 1 for the exclusively plant-based diet.

The mainly animal-based diet is built upon the Atlantic Diet (AD) guidelines. The AD is the traditional dietary pattern of Portugal and Galicia (Vaz Velho, Pinheiro and Rodrigues, Reference Vaz Velho, Pinheiro and Rodrigues2016) and it is considered a variation of the IMD, offering similar health advantages, especially on preventing cardiovascular diseases (Oliveira, Lopes and Rodriguez-Artalejo, Reference Oliveira, Lopes and Rodriguez-Artalejo2010; Guallar-Castillón et al., Reference Guallar-Castillón, Oliveira, Lopes, López-García and Rodríguez-Artalejo2013). It differs from the IMD for its higher intake of fish, red meat, milk, cheese, and potatoes, as observed in the findings of García-Gómez et al. (Reference García-Gómez, Rivas-Casais, Lorences-Touzón, Piedrafita-Páez, Muñoz-Ferreiro, Vázquez-Odériz and Romero-Rodríguez2022). This diet is included in the study, as it is quite similar to the IMD and people may shift to such dietary model with relatively low efforts, while contributing to reduce diet-related environmental impacts. For the AD, we assume the parameter β = 0.2, meaning a low resistance in changing the eating habits.

The mainly plant-based diet is based on the EAT-Lancet diet. Introduced in 2019 by the EAT–Lancet Commission, the EAT-Lancet diet is regarded as an exemplary model of a sustainable diet. This diet, as outlined by Willett et al. (Reference Willett, Rockström, Loken, Springmann, Lang, Vermeulen, Garnett, Tilman, DeClerck, Wood and Jonell2019), is not only health-conscious but also environmentally sustainable, playing a pivotal role in reshaping the global food systems while staying within the limits of planetary boundaries (Steffen et al., Reference Steffen, Richardson, Rockström, Cornell, Fetzer, Bennett, Biggs, Carpenter, de Vries, de Wit, Folke, Gerten, Heinke, Mace, Persson, Ramanathan, Reyers and Sörlin2015). The EAT-Lancet is a primarily plant-based diet, which also allows for a moderate consumption of meat and dairy products. It shares similarities with the IMD but incorporates a higher quantity of fruits and vegetables, whose consumption is low in the RID.

For the EAT-Lancet diet we assume a greater resistance to changing dietary habits toward a predominantly plant-based diet (β = 0.6). This suggests that people are assumed to be less inclined to adopt such dietary changes, reflecting the challenges associated with transitioning to a diet that places a greater emphasis on plant-based foods, such as social, religious, cultural, neophobic, and economic obstacles (Abe-Inge et al., Reference Abe-Inge, Aidoo, Moncada de la Fuente and Kwofie2024).

Moreover, we assess the WF for an exclusively plant-based diet. This choice aligns with prior research findings indicating that predominantly or entirely plant-based diets, such as vegetarian and vegan diets, exhibit minimal environmental impact, as supported by studies like those conducted by Hallström, Carlsson-Kanyama and Börjesson. (Reference Hallström, Carlsson-Kanyama and Börjesson2015), Chai et al. (Reference Chai, van der Voort, Grofelnik, Eliasdottir, Klöss and Perez-Cueto2019), and Cavaliere et al. (Reference Cavaliere, De Marchi, Frola, Benfenati, Aletti, Bacenetti and Banterle2023).

The exclusively plant-based diet avoids all animal-based foods, thus red meat, white meat, cured meat, fish, milk and cheese. We increase the minimum amount (in terms of grams per day) of pulses in order to compensate for nutritional lack of proteins deriving from the complete exclusion of animal-based products. Nevertheless, it's important to acknowledge that adopting a vegan diet may lead to nutritional inadequacies, and nutritional integration may be necessary to address the absence of nutrients like vitamin B12, essential fatty acids, and calcium (Alles et al., Reference Alles, Baudry, Mejean, Touvier, Peneau, Hercberg and Kesse-Guyot2019; Schupbach et al., Reference Schupbach, Wegmuller, Berguerand, Bui and Herter-Aeberli2017; Jeitler et al., Reference Jeitler, Storz, Steckhan, Matthiae, Dressler, Hanslian, Koppold, Kandil, Michalsen and Kessler2022). Our study prioritises optimizing the WF of the diets, addressing nutritional concerns is not the main focus. For the plant-based diet, we set β = 1.