Spatial-Temporal Footprints Assessment and Driving Mechanism of China Household Diet Based on CHNS

Food consumption is closely associated with resource consumption and environmental sustainability. An unreasonable dietary pattern would cause great pressure or damage to resources and the environment. It is particularly important to reduce the negative impact of household food consumption on resources and the environment while simultaneously ensuring people’s nutrient intake and health. This study applied the China Health and Nutrition Survey (CHNS) database to quantitatively study the spatial-temporal analysis of multiple footprints of household food consumption at multiple scales and explored the driving mechanism of the multiple footprints. The results showed that, except land footprint (LF), the other four types of footprints all decreased at varying degrees; the water footprint (WF), carbon footprint (CF), nitrogen footprint (NF) and energy footprint (EF) decreased by 18.24%, 17.82%, 12.03% and 20.36%, respectively, from 2000 to 2011; multiple footprints of food consumption of household in Guizhou was the highest among the 12 provinces involved in the study; this shows that resource consumption (water, energy and land resource) and environmental influences (CO2 emissions and nitrogen emissions) brought by food consumption of per household in Guizhou are much greater than in other provinces, which has a negative influence on sustainable development; by analyzing the driving factors of multiple footprints, it is shown that nutrient intake, household attributes, educational level and health conditions were significantly correlated to multiple footprints. Among them, nutrient intake has greater impact on the multiple footprints of Chinese household food consumption. By comparing multiple footprints of different dietary patterns, it was found that the current Chinese dietary pattern would cause excessive resource consumption, which would bring more pressure on resources and the environment. Adjusting household living habits would possibly reverse the unsustainable situation, such as reducing the consumption of animal-derived foods and adjusting the dietary pattern of households with a higher educational level and income status. Chinese Dietary Guidelines 2016 has better sustainability; the promotion of this dietary pattern across the country would help China to relieve the pressure on resources and environment from the consumer side, promoting the realization of sustainable development.


Introduction
Food consumption would have a profound impact on human health, resource consumption and environment sustainability [1,2]. With the rapidly growing population and higher living standards, food production due to increased demand would face greater pressure [3]. Agricultural production directly pressures water, energy, and land resources [4][5][6], and causes environmental issues including threats to biodiversity; increase in GHG

Research Area
China is the world's largest developing country and the world's second largest economy (China's GDP was USD 14.28 trillion in 2019, ranked 2nd in the world). China is also the most populated country in the world (1.4 billion in 2019) [22]. This study involved 12 provinces of China (Figure 1), which included Beijing, Liaoning, Heilongjiang, Shanghai, Jiangsu, Shandong, Henan, Hubei, Hunan, Guangxi Zhuang Autonomous Region, Chongqing and Guizhou as research areas. The population of these 12 provinces accounts for over 40% of the China's total population. The research included all the food consumption survey data that can be collected from CHNS, of 63,550 households, and the last five years provided by the database (2000, 2004, 2006, 2009, and 2011).

Food Classification
As is shown in Table 1, based on the statistics of the CHNS database, fifteen kinds of foods were divided into two categories: plant-based foods and animal-derived foods. Through elaboration of two-level food categories, among them, the primary classification included cereals and starchy foods; legumes and its products; vegetables and fruit; animalderived foods. In Chinese households, these are the major types of food consumed [21].

Methods
In order to quantify the footprint of the whole process of food consumption more accurately, this paper adopts the life cycle assessment method in the footprint assessment to calculate the footprints of plant-based and animal-derived foods at all stages of the process, from production to consumption, and establish a calculation model of multiple footprints [33]. The specific formulae are as follows: (1) Dietary WF: where P i represents the consumption of product i, g/cap/d, WF i represents the WF per unit yield of product i, and m 3 /kg, DWF represents the WF, m 3 /cap/d [34].
(2) Dietary CF: where P i represents the consumption of product i, g/cap/d, CF i represents the CF per unit yield of product i, and kgCO 2 eq/kg, DCF represents the WF, kgCO 2 eq/cap/d.
(3) Dietary NF: The NF includes nitrogen emissions in the whole life cycle of the food from production to processing, which can be divided into the NF of food production and food consumption [35].
where S protein i represents the amount of protein supplied by food in t; NC protein i is the nitrogen content of protein, which by default is 16%; W food represents the loss and waste of consumption in ton in the process; i represents different types of food.
VNF food i refers to the virtual nitrogen content of different foods and the virtual nitrogen content discharged into the environment during the process from production to consumption, which mainly exists in the volatilization of chemical fertilizers, runoff, crop harvest loss, processing loss, loss of animal manure and urine, etc. [36].
(4) Dietary EF: where DEF represents the dietary EF, MJ/cap/d, P i represents the consumption of product i, g/cap/d, EF i represents the EF per unit yield of product i, MJ/kg.
(5) Dietary LF: where DLF represents the dietary EF, ghm 2 /cap/y, P i represents the consumption of product i, g/cap/d, LF i represents the LF per unit yield of product i, ghm 2 /t. Footprint intensity of nutrient intake is calculated as follows:

Driver Factor Screening and Analysis
To analyze the factors for the differences between dietary WF, CF, NF, EF and LF at a household scale in China, and to compare the influence degree of these factors, 65 independent variables of 6 categories were selected from the CHNS database (as shown in Table 2). They are household attributes (V1 to V12), nutritional intake ratio (V13 to V16), labor and income status (V17 to V29), health and medical conditions (V30 to V49), educational level and social life (V50 to V54), and living habits (V55 to V65). The 65 proposed independent variables which were screened out and a correlation analysis was performed with multiple footprints and footprint intensities of nutrient intake at a household scale. The Spearman coefficient was calculated, and the significant variables were screened by significance test.

Data Sources and Processing
As is shown in Figure 2, the study was based on the food consumption survey data of 63 [38]. The multiple footprints of household food consumption were calculated.
The nutrition content of the edible part of food unit was obtained from the Chinese Food Composition Table [39,40] of the Institute of Nutrition and Health, Chinese Center for Disease Control and Prevention, and four nutrients including energy, protein, cholesterol and insoluble dietary fiber were selected to calculate the intensity of nutrient footprint in household food consumption. The independent variables of dietary footprint, nutrients footprint intensity and driving factors were imported into SPSS 25.0 for correlation analysis.

Spatial-Temporal Analysis for Multiple Footprints of Food Consumption
As shown in Figure 3, the dietary WF of China's household decreased from 4.33 m 3 /cap/d to 3.54 m 3 /cap/d, by −18.24%. Dietary CF decreased from 2.75 kg CO 2 eq/cap/d to 2.26 kgCO 2 eq/cap/d, by −17.82%, the maximum value for both of these factors were obtained in 2000. In addition, the NF decreased from 6.65 × 10 −2 g/cap/d to 5.85 × 10 −2 g/cap/d, by −12.03%. The EF decreased from 10.56 MJ/cap/d to 8.41 MJ/cap/d, by −20.36%. The LF showed the increasing trend first and then decreased, decreasing from 7.91 × 10 −4 ghm 2 /cap/d to 7.60 × 10 −4 ghm 2 /cap/d, which reached the maximum value of 8.18 × 10 −4 ghm 2 /cap/d in 2004. However, the change was not much obvious. Except LF, footprints produced by the food consumption of animal-derived foods were mostly higher than plant-based foods. During 2000-2011, the proportion of WF in plant-based foods decreased from 26.87% to 24.96%, the proportion of CF in plant-based foods decreased from 20.91% to 18.64%, the proportion of NF in plant-based foods decreased from 28.89% to 25.86%, the proportion of EF in plant-based foods decreased from 52.63% to 47.33%, and the proportion of LF in plant-based foods decreased from 67.71% to 65.31%. During 2000-2011, the proportion of WF, CF, NF, EF and LF in animal-derived foods increased from 73.13% to 75.04%, 79.09-81.36%, 71.11-74.14%, 47.37-52.67% and 32.29-34.69%.  This study also considered the differences among 12 provinces during the year of 2011. Whereas the WF of Henan households was observed to be the lowest (2.97 m 3 /cap/d), Guizhou was the highest (5.33 m 3 /cap/d), 1.79-fold higher than Henan. The EF of Chongqing household was recorded as the lowest (6.83 MJ/cap/d), while Guizhou was found highest (11.69 MJ/cap/d), 1.71-fold higher than Chongqing. Furthermore, the dietary CF and NF of Beijing households was the lowest, reached 1.85 kgCO 2 eq/cap/d and 4.35 × 10 −2 g/cap/d. However, the CF of Guizhou was the highest (3.19 kgCO 2 eq/cap/d), 1.72-fold higher than Beijing. The NF of Guizhou household was the highest (7.90 × 10 −2 g/cap/d), 1.82-fold higher than Beijing. The LF of Shanghai household was the lowest (5.78 × 10 −4 ghm 2 /cap/d), while Guizhou reached the highest (1.30 × 10 −3 ghm 2 /cap/d), 2.24-fold higher than Shanghai. More spatial-temporal analysis and composition characteristics for multiple footprints of 12 provinces were shown in Figures S1-S6 in Supplementary Materials.

Spatial-Temporal Analysis for Multiple Footprints Intensity of Cholesterol Intake
In Figure 7, EFI c changed the most during 2000-2011, decreasing by 11.31%. LFI c increased by 7.34%. The WFI c and EFI c of households in Shandong in 2011 was the lowest, 5.20 × 10 −3 m 3 /mg and 1.18× 10 −2 MJ/mg, respectively. The CFI c and NFI c of households in Beijing was the lowest, which was 3.10 × 10 −3 kgCO 2 eq/mg, 7.31 × 10 −5 g/mg, respectively. The LFI c of households in Shanghai was the lowest (0.93 × 10 −6 ghm 2 /mg).

Driving Factors of Multiple Footprints of Food Consumption
In Figure 8, most of the proposed independent variables showed different degrees of correlation with multiple footprints and multiple footprints intensity of nutrient intake. The nutrient intake illustrated a strong correlation with multiple footprint and nutrient intake intensity. In Figure 8a, among all the correlations, the proportion of protein intake (V16) and WF presented the strongest positive correlation, while the proportion of carbo-hydrate intake (V13) and WF had the strongest negative correlation. It can be found that the driving factors of WF, CF and NF are consistent to some extent. Among the 65 proposed independent variables, 53 (81.54%) were significantly correlated with the WF, while 46 (70.77%) correlated with the LF. In Figure 8b, it illustrated that the strongest positive correlation was found between proportion of fat intake (V14) and WFI e , and a strong negative correlation was found among proportion of carbohydrate intake (V13) and WFI e . A certain degree of consistency was observed among the driving factors of WFI e , WFI p and WFI f , where 49 (75.38%) of the variables correlated with WFI c , and 43 (66.15%) correlated with WFI p .  Figure 8c shows that the strongest positive correlation was found among the proportion of protein intake (V16) and CFI e , and negative correlation was obtained among carbohydrate intake (V13) and CFI e . Among these, 49 (75.38%) correlated with CFI f , and 43 (66.15%) negatively correlated with CFI p . Figure 8d shows a strong positive correlation between protein intake (V16) and NFI e , while there was a negative correlation between carbohydrate intake (V13) and NFI e . It can be seen that 53 (81.54%) variables correlated with the NFI e , and 42 (64.62%) correlated with NFI p . According to Figure 8e, a strong positive correlation was observed between proportion of carbohydrate intake (V13) and EFI c , the strongest negative correlation was observed between protein intake (V16) and EFI c . Overall, 52 (80.00%) variables correlated with EFI c and 43 (66.15%) correlated with EFI p . In Figure 8f, among all the correlations, the strongest correlation was province (V1) and LFI p . There were 53 (81.54%) proposed independent variables that correlated with LFI c , and 40 (61.54%) correlated with LFI e .

Scenario Analysis of Different Dietary Patterns
This study selected China's food consumption data during 2011 as the basic scenario (baseline). These results were compared with two developed countries: the United States (S1) and Japan (S2). The United States was selected because it is the most developed country [41], while Japan was selected because it borders China and has a similar household dietary habit to China [42]. Additionally, more reliable data from 2015 can be obtained for the United States and Japan. The study also chose Chinese Dietary Guidelines 2016 (S3) [43], recommended food consumption of Guideline of Australia 2013 (S4) [44], and the food consumption guidelines issued by Germany 2013 (S5) as reference scenarios [45].
The results from Figure 9 demonstrated that the CF, EF and LF of China's food consumption pattern during 2011 were highest among all scenarios. The WF and NF dietary pattern recommended by China were found to be the lowest among several scenarios. When the dietary pattern reaches the recommended value of Chinese Dietary Guidelines 2016, the WF, CF, NF, EF and LF would be reduced by 56.48%, 69.47%, 43.57%, 47.44%, and 54.91%, respectively. The WF, NF and LF of plant-foods consumed by Chinese households in 2011 were the highest among all scenarios, and the CF, EF and LF of animal-derived foods consumed by Chinese households in 2011 were also the highest among all scenarios. Meanwhile, the biggest reduction in the consumption of plant-based foods would be the WF, which would reduce by 31.14%, and the biggest reduction in the consumption of animal-derived foods would be the CF, which would reduce by 75.53%. Due to the different dietary habits, there were significant differences in nutrient intake among households in different countries. As can be seen from Figure 10 In 2011, the dietary protein intake of Chinese households was higher than that of the United States and Japan, but the intake of high-quality protein from aquatic products and other animal-derived foods was lower. However, the intake of nutrients under the recommended recipes in Australia and the dietary guidelines in Germany was relatively low, and the footprint value was also relatively low. The intake of nutrients was in line with the dietary pattern recommended by the Chinese Dietary Guidelines 2016, which could be used as a reference.

Driving Mechanism of Household Dietary Footprints
Results show that multiple footprints of household food consumption have strong correlation with household nutrient intake. The water footprint of animal-derived foods is higher than plant-based foods, consequently presenting greater impacts, particularly on water footprint [29]. It can be seen that household animal-derived protein and cholesterol intakes increased with the proportions of animal-derived foods in household dietary pattern. This trend might increase the pressure on shortages of water resources [46]. From 2000 to 2011, the dietary energy of urban and rural residents in China mainly came from cereals (or carbohydrates); the protein mainly came from cereals [47]. Additionally, the results showed that the proportion of carbohydrate intake was negatively correlated to multiple footprints. To a certain extent, it indicates that a dietary pattern partial to carbohydrate intake is more environmentally friendly than a dietary pattern partial to high animal protein and cholesterol intake. Previous research demonstrated that each person should reduce their consumption of animal-derived foods by 205.1 kg CO 2 , equivalent to 12.1% per year [48]. Other studies reported that, in order to support health and achieve the climate stability goal, beef could be replaced with pea protein to reduce the environmental footprint of animal-derived foods [49,50]. Greenhouse gas (GHG) emitted from the production of plant-based beverages (such as oats, soybeans, almonds and rice milk) contributes only 22-38% to the total greenhouse gas emitted from the milk production. Thus, replacing milk and other dairy products with plant-based beverages could also greatly reduce water consumption [51].
Footprints of household food consumption shows a strong correlation with educational levels. The proportion of carbohydrate, protein and fat intake had changed with the change of household dietary pattern in China, which led to a great impact on water and carbon footprint. As shown in the study of high carbohydrate intake among adult women in Botswana, educational level may influence the choice of food intake [52]. The increase in education level would directly increase the consumer's attention to the rationality of nutrient intake, and tended to increase the intake of protein instead of calories [53]. The educational level also affects the household choice of proportion of animal-derived and plant-based foods [54], which directly affects household intake of carbohydrate, protein and energy, and indirectly drives the change of multiple footprints of household food consumption.
With the increase in income status, people tended to increase nutrient intake. However, people with a higher income status tends to increase the consumption of more refined foods, price level and taste grade [55]. The increasing complexity of food processing also increased the multiple footprints of food. Due to different income status of household in different provinces, or urban and rural areas, variations among household eating habits are formed. A previous study argued that the more developed and urbanized a household, the more likely it was to have a higher sugar, fat, and highly processed and packaged food intake [56]. Hence, rising income and urbanization effectively drives the dietary transition, where the traditional diets are replaced by diets with more refined sugars, refined fats, oils and meats [57]. Households of developed provinces increase the consumption of various animal-derived foods (particularly poultry and pork), while in rural areas, households increase the consumption of pork. Study shown the urban and rural households have no strong preference to beef and mutton. Rural households who increase the consumption of poultry, beef and mutton will reduce their consumption of pork and increase consumption of other animal-derived foods [58]. These results illustrated the differences of household food consumption footprints between urban and rural areas. Therefore, provinces and urbanization are the key factors for driving of household food consumption footprints.
The convenience of public and private transportation reduced the calorie consumption of households, and high-fat and high-protein foods become cheaper. Urbanization promoted household incomes together with the increasing consumption of such foods. People can get more expensive calories from non-starch foods, so they choose to reduce the consumption of rice and flour [59]. Currently, China is experiencing the transition from a developing to a developed country, having the influence of huge population and trend of developing economic, household food waste became serious issue, resulting in the unnecessary consumption of resources [60].
Among 12 provinces involved in the study, the results have shown that the multiple footprints of food consumption were highest for households in Guizhou. Guizhou is rich in natural products and diverse in diet. The terrain in Guizhou is complex and mainly mountainous. The mild and humid climate also brings Guizhou an advanced planting industry [61]. As shown in our study, the beef consumption of household in Guizhou (35.57 g/cap/d) was not higher than most provinces. However, the multiple footprints of unit yield of beef in Guizhou were the highest among the 12 provinces. Thus, the WF (0.66 m 3 /cap/d), CF (1.33 kgCO 2 eq/cap/d), NF (3.20 × 10 −2 g/cap/d) and LF (0.20 × 10 −4 ghm 2 /cap/d) of beef consumption of household in Guizhou is the highest among 12 provinces, while the EF (0.33 MJ/cap/d) is also higher than most of the provinces. This situation also appears in several other animal-derived foods. Moreover, because the multiple footprints of per unit yield of animal-derived foods are higher than that of plant-based foods, the dietary multiple footprints of households in Guizhou are higher than that of other provinces. In order to achieve sustainable development, it is suggested that Guizhou should reduce their consumption of animal-derived foods by utilizing the unique Karst landform and the potential of the grassland to develop its animal husbandry [62].

Suggestions for Sustainable Dietary Adjustment
Based on the above, the dietary pattern of households in China is not a sustainable form of development; the following suggestions are proposed: (i) we suggest that households should reduce the consumption of red meat (mutton, beef, etc.), because the multiple footprints of unit yield for red meat is higher than other kinds of foods (e.g., the WF of mutton was 14.55 m 3 /kg, 92.46 times higher than that of vegetables). We also suggest that households should consider conforming to a balanced diet, reducing the excessive intake of food and keeping food waste to a minimum. Based on the increasing demand for animal food, we suggest using plant-based foods rich in protein as a substitute for some animal-derived foods, and swap refined grain choices for whole grains, which not only meet household requirements for nutrient intake and health keeping, but also achieve the purpose of reducing the consumption of dietary resources [49,50,63]. Residents should be encouraged to get protein through beans, using plant-based beverages instead of milk and other animal dairy products [51]. Food waste, reduce resource consumption and greenhouse gas emissions caused by the loss of food production and consumption [64] should be reduced. We also suggest that people with higher income and educational level adjust their dietary pattern according to dietary guidelines issued by China (2016) [43], avoiding the blind pursuit of the high-sugar and high-fat dietary pattern common in Western or developed countries, reduce excessive intake of nutrients, and choose foods with the same nutritional value and lower resource footprint [53]; (ii) we also suggest that the Chinese Government encourages the development of sustainable production and processing technology in the agriculture and food processing industry, and improve household awareness of dietary guidelines. Additionally, we suggest that the Chinese Government pays more attention to guiding consumers, such as publicity and education, and spreads the use of footprint calculators or footprint price lists for households, making dietary choices clearer; (iii) we suggest that nutrient research institutions consider household nutrient intake alongside resource consumption when formulating recommended dietary pattern.
Future studies should focus on the social response to dietary multiple footprints and methods of reducing resources consumption and GHG emissions globally. Further improvements in data integrity with better classification of foods and broader survey are needed to provide a more sophisticated understanding of the influences of diet on environment and sustainable development. A more comprehensive understanding of the impact that the driving mechanism of household dietary has on the multiple footprints on scales will facilitate households, governments and research institutes to adjust dietary patterns, accelerate technological innovation and further studies on the relationship between food consumption and environmental impact, alleviating the current resource and environmental pressures to achieve absolute sustainability.

Conclusions
This paper studied the spatial-temporal characteristics of multiple dietary footprints and analyzed its driving mechanisms. The main conclusions were as follows: (1) During 2000-2011 in China, household dietary consumption WF, CF, NF and EF were decreased by 18.24%, 17.82%, 12.03% and 20.36%, respectively. However, the LF remained stable within the time scale of the research. For footprint intensity, the EFI e (−11.52%), EFI f (−9.74%) and EFI c (−11.31%) have decreased, while LFI p , and LFI c have increased by 10.23% and 7.34%, respectively. (2) Nutritional intake ratio, household attributes, educational and health consciousness had a stronger correlation with dietary footprints. Among these, protein intake had a stronger positive correlation with WF, NFI e and CFI e , while the proportion of carbohydrate intake had a stronger negative correlation with WF, WFI e , CFI e and NFI e . The proportion of fat intake had a stronger positive correlation with WFI e . The proportion of carbohydrate intake had a stronger positive correlation with EFI c , and the proportion of protein intake that had a stronger negative correlation was EFI c . (3) The multiple footprints of Chinese household food consumption are much higher than the dietary pattern recommended in the Chinese Dietary Guidelines 2016. It is strongly suggested that the households reduce animal-derived foods with high footprints (especially beef and mutton). Additionally, household should increase the intake of fruits and vegetables to reduce the size of the multiple footprints.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.