Household perceptions of factors that affect food consumption in grassland areas: a case study in the Xilin Gol Grassland, China

Food consumption acts as an intermediary that connects ecosystems with human systems in grassland areas. In this study, we used fuzzy cognitive mapping to quantitatively assess the factors that affect food consumption in the Xilin Gol Grassland, China, from the perspective of local rural households. We found that household perceptions of the factors that affect food consumption differed among parts of the grassland transects in both the number and the strength of these factors. Livestock numbers, household income, regional economic development, consumption habits, age, and infrastructure were the most important factors mentioned by the farmers and herders, but herders were more sensitive to ecological and economic factors, whereas farmers focused more on personal and social factors. The differences in the main factors between study areas revealed the key economic, social, and ecological dimensions. Our results provide a reference for policymakers to develop improved policies to encourage regional sustainable development.


Introduction
Food consumption is essential for human survival, but perceptions of this consumption vary among regions and cultural backgrounds. In grassland areas, food comes directly from the grassland areas and the food supply is limited [1,2]. Food consumption depends on the ecosystem's provision of services, and is determined by the relationship between ecosystem services and human activities [3,4]. However, in grassland areas, the ecosystem is often very vulnerable and the food consumed by local residents comes primarily from that ecosystem, which creates pressure on the local ecosystem [5].
Moreover, socioeconomic development increasingly affects consumption patterns, and often leads to increasing contradictions between human activities and local ecosystem needs, and these contradictions create challenges to regional development, especially in grassland areas [6,7]. In the sustainable development goals released by the United Nations [8], food consumption is linked to goal 2 (eliminating hunger) and goal 12 (responsible consumption and production). Consumption therefore plays an important role in sustainable development issues around the world [9]. Despite the importance of these goals, citizens of some developing countries have developed predatory ways of exploiting and using environmental resources during urbanization and socioeconomic development, and this has led to sustainability issues during regional development [10].
For example, northern China's Xilin Gol Grassland, one of the country's largest grassland areas, is located in the agro-pastoral transitional zone. Due to the grassland's sensitive and fragile environment and intensive human activities, this area faces serious challenges to sustainability [11,12]. The local government and China's national government have invested more than 2 × 10 8 RMB for regional ecological projects [13]. Policies such as the Grassland Ecological Compensation Policy and the Beijing-Tianjin Sandstorm Source Project that have been implemented in this area have greatly affected the lives and production of local residents [14,15]. This, in turn has affected the daily food consumption of residents. For example, because of the strict ecological policy, local herders reduced their family livestock number and increased the percentage of cows; this affected the family meat supply, leading families to eat more beef than mutton, in contrast with the previous dominance of mutton [9]. Moreover, residents of this area come from many national backgrounds, including Han Chinese, Mongolian, Hui, and Man [13], and each of these groups has different cultural preferences. The grassland's cultivated land is mainly distributed near its southern border, and the limited food supply provided by this land cannot meet all regional needs [16]. Thus, the issues related to food consumption in this area are complex, and this suggests equally complex household perceptions of the factors that affect their food consumption.
Several researchers have studied food consumption and the factors that affect consumption in rural regions [17][18][19]. However, there have been few studies in pastoral regions and agro-pastoral transitional zones [9], especially in Inner Mongolian of China. Studies of the factors that affect food consumption have mainly focused on economic, social, and ecological factors. The economic factors include household income, the level of economic development (e.g. GDP), economic differences between urban and rural areas, prices of food and other necessities, and access to markets [20][21][22][23][24]. The social factors include problems related to an aging population, urbanization, cultural diversity, and cultural practices such as festivals that require specific foods [25][26][27][28]. The ecological factors include the time of year, land-use diversity, climate, and many environmental factors [9,26,29]. The main food consumption impact factors in different areas are different even the different transitional zones. The agro-pastoral transitional zone in China is from northeast China to Tibet. Research showed that in Northwest China household food self-supply and household income affected family food consumption pattern [30]; in Tibet household food consumption was affected by policy and transportation accessibility [31,32]; in Inner Mongolia household food consumption was relative to culture, family self-supply and income [9,26], etc. Statistical data, questionnaire responses, and remote sensing data are the most commonly used data in studies of the factors that affect food consumption [17,[33][34][35]. Several study about stakeholder perception conducted before and revealed people's concerning indicator for fresh food production [36], meat industry [37], food risk [38] and so on.Nevertheless, there have been insufficient studies of stakeholder perceptions of these factors, particularly in the Xilin Gol Grassland, where we found rare qualitative and quantitative analyses of these factors.
Fuzzy cognitive mapping is a semi-quantitative modeling tool that provides a framework for comparing knowledge obtained from non-technical experts [39,40]. It can help to reveal how social groups such as farmers and herders think about an issue and allow researchers to build a network model of a situation or system [41]. It has been successfully used in the fields of geography, social science, ecology, and economics [42][43][44]. Such studies are easy to conduct and effective at revealing complex and subjective beliefs [45][46][47]. They are particularly useful during preliminary research to identify factors and relationships that should be examined in more depth in future research.
In this study, we used primary data obtained from interviews with residents of the Xilin Gol Grassland to support fuzzy cognitive mapping, and combined this data with regional land-use and socioeconomic data to identify the factors that affect food consumption in this region. Our objectives were to use the perceptions of local households to explore these factors, identify the key impact factors, and reveal differences in these factors among residents of the grassland. We focused on these factors to help us understand the coupling relationships between the region's human systems and its ecosystems, and to provide empirical support for future regional development and the associated policy development.

Study area
The Xilingol League lies in the central part of China's Inner Mongolia Autonomous Region. It belongs to the Mongolian Plateau, with an elevation that ranges from 800 to 1800 m asl. Annual precipitation totals 288 mm, and annual pan evaporation ranges between 1700 and 2600 mm. The annual average temperature is 3.60 • C, with mean monthly temperatures ranging from −20 • C in January to 21 • C in July. The league comprises two county-level cities, one county, nine banners, and two districts. The total registered population was 1.036×10 6 [48].
Most of the Xilin Gol Grassland is located within the Xilingol League. The grassland cover can be divided into typical grassland, desert grassland, cultivated land, and sparse grassland (figure 1). The carrying capacity of these grassland types gradually decreases from south to north.
In this study, we selected three typical areas of the Xilin Gol Grassland for investigation: the West Ujimqin Banner, which is a typical pastoral grassland area; the Zhenglan Banner, which is a pastoral area of the Hunshandake Sandy Land; and the Taibus Banner, which is an agro-pastoral transitional zone (figure 1). The three areas represent a north to south transect through changing environments and ecosystems. The West Ujimqin Banner is in the eastern part of the Xilingol League, whereas the Zhenglan Banner and Taibus Banner are located in the southern part of the league. All three banners are in an arid to semi-arid transitional zone, with annual precipitation around 400 mm, and the annual pan evaporation is very high (>1800 mm). The ecosystems in these study areas are very vulnerable to degradation (table 1).

Data collection 2.2.1. Interviews and data collection
We interviewed representative local residents living in the three rural areas of the Xilingol League from 20 July to 5 August 2019. Details of our sampling are provided later in this section. Data were generated by means of fuzzy cognitive mapping during face to face interviews. The interviewees were one person per family who understood the family's food consumption and other key details such as their livelihood, education, and so on. Food consumption types of rural people in the Xilin Gol Grassland mainly included grains, vegetables, fruits, meat, eggs, milk and non-staple food [9]. In this study, we did not use a standardized questionnaire to guide the interviews, since our goal was to discover the factors that interviewees considered important rather than their responses to preselected factors that we considered important. Instead, we asked a simple question: 'What factors affect your food consumption?. ' We showed an example of a fuzzy cognitive map to the interviewees before we started the interview. We then asked them to define the factors that they believed would affect their food consumption and to draw the relationships among these factors on a piece of paper. They assigned links between the factors and rated the strength of each link from −5 (i.e. increasing one factor strongly decreased the other factor) to +5 (increasing or decreasing one factor produced the same change in the other factor). The direction of each link was shown with an arrow. The larger the absolute value of the score, the stronger the impact. The networks were drawn by the interviewers in response to the instructions of the interviewees. Appendix A provides a typical example of a map. During the interview, interviewees could add or change the factors or their weights. The final version of each network was approved by the interviewee. We obtained personal details from the interviewees and promised that this information would remain confidential and would only be used for our scientific research. The personal details included the interviewee's name, household location, gender, vocation (farmer, herder, or both), age, education (number of years of school), family size, rangeland and cropland area managed by the household, ancestry, household income, and the interview date.
The interviews were undertaken in Mandarin Chinese or Mongolian, depending on the language preferences of the interviewees. Local people such as officials working at the local grassland observation stations, the head of the village, or young people who were college students acted as translators when this was necessary to ensure clear communication with each interviewee. About 50% of interviewees needed assistance from the translator. We conducted a total of 67 interviews.
The interviewees were chosen by stratified random sampling. First, we defined the scope of the survey based on the typical land use categories in each of the three study areas. Second, we identified the study sites using stratified random sampling based on whether the site was dominated by herder or farming households. We interviewed households in 5 towns or sumu (sumu is the town-level administrative division used in the Inner Mongolia Autonomous Region) in West Ujimqin Banner, 4 towns or sumu in Zhenglan Banner, and 3 towns in Taibus Banner as our study sites. Appendix B summarizes their names and characteristics. We used pseudo-random sampling to choose the families that we interviewed at each site because the low population density (i.e. 0.2 person per km 2 ) led to a large distance between families, and required concentration of the families in some locations. We continued our surveys until no new factors were identified, thereby ensuring that the fuzzy maps adequately represented each interviewee's perceptions, which was proved to be effective in the previous studies [47].

Basic information about the interviewees
We interviewed 21 families in West Ujimqin, 21 families in Zhenglan, and 25 families in Taibus in 2019. Figure 1 shows the locations of the interview sites, and table 2 summarizes the demographic characteristics of the interviewees. The average age of the interviewees increased from north to south, accompanied by a decrease in the proportions of Mongolian families and female interviewees (table 2). The socioeconomic characteristics of the families also differed among the study areas.

Data analysis
Our 67 interviews revealed 28 different factors that directly affected food consumption (table 3). We classified the factors into four dimensions: ecological factors (10), social factors (5), economic factors (5) and personal factors (8). Here our classification referred to the usual sustainable development indicator and dimension [3], while the factors relative to the personal characteristic and personal emotion were sorted as the personal dimension. Four dimensions consisted of all types of the impact factors.
Each interviewee scored the strengths of the relationships among the factors in the fuzzy cognitive maps. These strengths (V) were divided into four degrees: very weak (0 < V < 1), weak (1 ≤ V < 2), important (2 ≤ V < 3), and very important (3 ≤ V < 4). Although we expected the scores to be between −5 and 5, in practice the interviewees only assigned values between −4 and 4. Consumption pattern means 'regional consumption behaviors (e.g. when people eat mutton, they drink alcohol; family usually stock and eat a whole sheep or beef in winter)' . Conformity means 'Conformity with neighbors or cultural traditions' .

Overview of the factors that affect food consumption
According to the interviewees, 28 factors affected the food consumption in their daily life (table 3). However, the factors differed among the study areas. In the pastoral areas (Zhenglan and West Ujimqin), interviewees identified more direct and indirect factors than in the agro-pastoral transitional zone (Taibus) ( figure 2(a)). However, the numbers did not differ significantly between the three areas, with an average of 7 direct impact factors and 4 indirect intermediate impact factors in each area. Among the four dimensions of the impact factors, personal and ecological factors were perceived as the most important factors in all three banners, with both a greater number of mentions ( figure 2(b)) and a greater cumulative score (figure 2(c)) than the other factors. The key impact factors analysis also indicated that interviewees showed stronger cognition in personal and ecological factors, which we will discuss in section 3.2.
From the view point of each banner, the factors in four dimension also showed the differences. In Taibus Banner, ecological (8 times per interviewee) and personal factors (7 times) were mentioned most often and social factors were mentioned least often (3 times). In Zhenglan Banner, personal factors were mentioned most frequently (10 times), and economic factors were mentioned least frequently (4 times), although economic factors were ranked 3rd instead of 4th in the cumulative scores. In West Ujimqin Banner, the pattern was similar to that in Zhenglan, which means that even though residents did not mention economic factors frequently (5 times), they ranked the economic factors highly. Although Taibus residents did not mention ecological factors and personal factors as often as residents in Zhenglan and West Ujimqin did, they gave both dimensions higher scores. Residents in Taibus considered the social factors to be least important, with the fewest mentions (3 times) and lowest score (1.8 in total). This may be because socioeconomic development in the pastoral areas (Zhenglan and West Ujimqin) has been very rapid since 2008. We will discuss this in more  detail later in the paper. The residents of these areas strongly perceived the changes and the impacts on their lives, which would lead them to focus more on this dimension.
Our analysis of the structural and functional demographic variables showed that residents in Taibus, Zhenglan, and West Ujimqin banners differed significantly in their perceptions of impact factors, family size, rangeland size, cropland size, number of sheep, number of cows, years of education, household income, and subsidies (table 4).

The key impact factors and their characteristic
Livestock numbers, household income, regional economic development, consumption habits, age, and infrastructure were the most important impact factors mentioned by the interviewees and showed the most interactions with other factors ( figure 3). In all study areas,the high scores that interviewees gave goes to livestock numbers, household income, regional economic development, consumption habits, which were all showing V > 2.0 (important degree). Especially, age was thought as an important factor affected household daily food consumption in Taibus, where V = 2.6, showing important degree compared to other two areas). Besides, though the social factors were ranked lower than other factors, infrastructure were scored highest among all the social factors in West Ujimqin (V = 1.8) and Zhenglan (V = 1.2), which also showed distinguishing difference between pastoral areas and agro-pastroal transitional zone (V = 0.1 in Taibus). Thus, in this paper we focus on these important factors and analyzed their characteristic.
Additionally, through the key impact factors we found that interviewees paid more concerns on the ecological factors and economic factors, which could also find the clue s in figure 2interviewees showed high frequency and high cumulative score in ecological factors, while relative low frequency and high cumulative score in economic factors. It suggested that ecological factors likes livestock numbers are so important that interviewees all mentioned them frequently as well as gave high score; though interviewees mentioned economic factors not so often, they tended to gave high score as long as they gave the economic factors a nomination.
Among all these key factors, interviewees were more concerned about ecological factors (livestock number) in the northern areas (Zhenglan and West Ujimqin) than in the southern area (Taibus) ( figure 3). In addition, personal factors (especially consumption habits) and economic factors (especially household income and regional economic development) were ranked as the second and thirdmost important groups of factors after the ecological factors in the pastoral areas (Zhenglan and West Ujimqin), whereas in the agro-pastoral transitional area (Taibus), ecological factor (livestock number) was also ranked first, and economic factors (especially household income) were ranked as the secondmost important factors. These results suggest that as we moved from the north to the south, residents became more concerned about the economic and personal factors and less concerned about the ecological factors.

Ecological factor: number of livestock
In all three study areas, the number of livestock was considered the most important factor (V > 3 in the three study areas) that affected daily food consumption ( figure 3). The number of livestock directly affects a resident's food consumption by defining the size of this food resource, but also affects food consumption indirectly by its effect on household income. This was true in all three areas.
However, we also observed a trend from north to south, in which the number of livestock was affected by a change in priorities from ecological factors (number of livestock, season, rangeland) and economic factors (household income, regional economic development) to economic factors (household income, regional economic development, food price) and personal factors (consumption habits, age, family structure). Especially in the pastoral areas (West Ujimqin and Zhenglan), the number of livestock was strongly affected by the rangeland area. Also, the number of interviewees who emphasized the number of livestock tended to decrease moving from north to south.

Personal factors: consumption habits and age
The most important personal factor involved consumption habits (V > 2.5 in all three study areas). In all three study areas, this was ranked as one of the most important factors, and one that interacted strongly with the other factors (figure 3). These habits were mainly affected by household income, vocation, and ancestry, and all of these factors were also affected by regional economic development. Even though residents' consumption behaviors have probably changed over time in response to other changes in their lives (e.g. rapid economic development), these changes were reflected in their consumption habits at the time of our interviews.
Age had a particularly important impact on food consumption in Taibus (V = 2.6), which relates to the problem of an aging population in this area. Residents' average age in Taibus was 62.9, which is more than 10 years older than in the other regions (table 2). In addition, the areas of rangeland and cropland for each family were smaller in Taibus than in the two pastoral areas (table 4), and interviewees considered their income from agriculture too low to provide a comfortable life for young people; as a result, it was mostly old people who chose to stay in rural areas in Taibus. The aging problem affected consumption directly by leading residents to eat less and restrict their food choices, and indirectly by affecting their health, income, vocation, and residence location (figure 3).

Economic factors: regional economic development and household income
Regional economic development is one of the most important economic factors that affected other factors that in turn affect food consumption (V = 2.5 in West Ujimqin, V = 2.3 in Zhenglan, V = 2.3 in Taibus) (figure 3). In the pastoral areas, economic development improved the infrastructure and increased household income, thereby changing ways of life and affecting food consumption. In the agropastoral transitional zone, interviewees believed that their nutrition had improved and that their number of livestock had increased, making herding more profitable, and this affected their choice of vocation and food consumption.
Household income was another important economic factor that interacted with many other factors and differed among the three study areas (V = 2.4 in West Ujimqin, V = 2.3 in Zhenglan, V = 2.9 in Taibus) ( figure 3). In the pastoral areas (Zhenglan and West Ujimqin), household income was mainly affected by factors related to animal husbandry such as the number of livestock, cost of production, food price, food market maturity, agriculture policy, and area of rangeland. In the agro-pastoral area (Taibus), household income was mainly affected by the personal and social factors (i.e. vocation, health, family help, family structure).

Social factor: infrastructure
In the pastoral areas, infrastructure had the highest score among the social factors (V = 1.8 in Zhenglan, V = 1.2 in West Ujimqin), but in Taibus, it had a much weaker effect on food consumption (V = 0.1; table 4, figure 3). Due to the large grassland area, the infrastructure (mainly roads) was poor and limited residents' access to markets and the outside world in the pastoral areas. Along with economic development and government policy, the infrastructure improved, thereby improving food consumption and offering more diverse food choices that became easier to access. However, in the agro-pastoral area (Taibus), which had a higher population density and a smaller area, the cost for governments to build basic infrastructure was lower and rural residents found it easier to get access to local markets where they could obtain more food, thereby weakening the strength of the social factors in this area (figure 3).

Key factors that affected food consumption
Based on changes in the socioeconomic characteristics of the study areas from 2000 to 2017 and the basic information on farmers and herders that we obtained in the 2019 field survey, we analyzed the social, economic, and ecological dimensions of the study areas to reveal causes of the differences in the number of livestock, household income, regional economic development, consumption habits, and age in the three typical areas based on the fuzzy cognitive mapping ( figure 3).
The number of livestock differed greatly among the three study areas. The number of livestock in the study area also decreased (in response to a government policy to prevent overgrazing) at different rates from 2000 to 2017. In 2017, the number of livestock in West Ujimqin was 930 900, versus 365 600 in Zhenglan and 191 200 in Taibus ( figure 4(c)). The residents of West Ujimqin tended to herd more sheep, whereas residents of Zhenglan tended to herd more cattle, and residents of Taibus tended to herd a small number of animals with a high economic value, such as camels and horses. In addition, the size of rangeland decreased moving from north to south (table 4) figure 4(a)). The household disposable income of residents in West Ujimqin (23 662 yuan) was always highest, and the income decreased from north to south ( figure 4(b)). Moreover, the GDP differences between the three areas widened during the study period, which suggests there were significant regional differences in the economic conditions. The number of livestock decreased from 2000 to 2017 in West Ujumqin and Zhenglan, but remained relatively constant (despite some fluctuation) in Taibus. Besides, residents in Xilingol League witnessed a significant wealth growth in the past decades, on one hand it was mainly contributed by the resource exploitation (i.e. mining) [6], which also posed challenges to regional sustainable development; on the other hand, the rapid economic development would change residents' daily food budget as well as daily food consumption structure [16] , which raised new demand for regional food supply.
Consumption habits and age also differed significantly between the three regions. Previous research on our study areas showed that the food consumption by residents of the three study areas differed greatly in quantity, structure, and preference. In West Ujimqin, residents tended to consume meat, vegetables, and fruits; in Zhenglan, residents tended to consume meat, grains, vegetables, and fruits; and in Taibus, residents tended to consume grains, vegetables and fruits (table 1). This suggests changes in the diet from primarily pastoral to primarily agricultural areas.
The mean age of residents of the three study areas increased from north to south, from 45 years in West Ujimqin to 51 in Zhenglan and 63 in Taibus. However, the total population of each region remained stable from 2000 to 2017 ( figure 4(d)). However, the age structure in Taibus represents a potential problem. On the one hand, the banner had a much higher population than the other two regions (figure 4d) as well as a high population density (62 persons per km 2 ; table 1, figure 4d); on the other hand, large numbers of rural people have moved into the banner's cities to find jobs, leaving a growing proportion of elderly people in rural areas. Many of the elderly people are unable to find jobs in the city or preferred to remain in rural areas due to the low cost of living.
As GDP increased, infrastructure development also increased, especially in terms of the construction of roads. In Taibus, the small land area (table 1) means that construction of roads has not caused problems related to access to food supplies; that is, a relatively small number of roads can meet the needs of residents. However, in West Ujimqin and Zhenglan, which are larger, governments have had more serious problems building sufficient roads.

Differences in household perceptions
In the Xilin Gol Grassland, there is a natural gradient of ecosystems and environmental conditions from north to south. Our results and previous research suggest that this gradient has affected the local culture, customs, and perceptions of local residents [14,26,49]. The socioeconomic characteristics of the residents differed among the three areas, as did their perceptions of the grassland system that sustained them, with both direct and indirect impacts on human systems and activities in the Xilin Gol Grassland. Human activities have adversely affected the region's vulnerable environment; on the one hand, overgrazing and unsustainable use of local natural resources (i.e. grassland, water) led to the environmental issues [6,7]; on the other hand, since the implementation of key ecological projects (i.e. Grassland Ecological Compensation Policy, Grain for Green) by the government, the local ecosystem has recovered since 1999 [50,51]. As we moved from north to south along our transect, residents relied less on grasslands for their living and more on agriculture. Thus, their perceptions of the importance of the grassland weakened, and they showed less concern about the ecology of their local grasslands. Such transitions along a grassland transect have been described previously [14,52]. Furthermore, in the other grassland areas, people's food consumption behaviors were found to be affected not only by infrastructure, consumption habits, household income, local food supplies, and family size [18,53], but also market accessibility, eco-friendliness, local production, and nutrition [53][54][55].
Moreover, household perceptions between Han people and Mongolian people showed some differences in the impact factor scale and score of factors (appendix D1). We took Zhenglan as example to divide the interviewees into two group by nationality of the family members, in which Han family group was consist of 11 interviewees and Mongolian family group was consist of 10 interviewees. The results showed that Mongolian interviewees perceived more factors (26) than Han interviewees (18). What's more, Mongolian showed special focus on infrastructure (V = 2.0 ) while Han interviewees especially concerned the season (V = 2.7).It suggests that the infrastructure (e.g. road) did affect household food consumption a lot through the food accessibility; while the Han family tend to plant some vegetable in the yard which are one of most important sources of food, and the planting of vegetables is directly relative to the season in this area.

Limitations of our study
The largest limitation on our study related to budget and time constraints, which limited the number of residents we could interview. In order to test the effectiveness of our results in limited sample scale, we divided all the family into two group randomly to compare the results in three banners. The results showed that there were not significantly difference between two groups in all three banners (appendix D). On one hand, no matter in West Ujimqin (appendix D2), Zhenglan (appendix D3) or Taibus (appendix D4), the scale of food consumption impact factors revealed by the interviewees were close; on the other hand, the high scored factors and high frequency factors in different groups within the banner were similar and kept consistency to the results of the whole banner.
We nonetheless found interesting differences among the three banners that suggest a need for additional research to clarify these differences. Fuzzy cognitive mapping provided an effective way to analyze household perceptions of the factors that affected food consumption and to provide a quantitative (though subjective) description of the study area. However, there are limitations of this approach, including the subjectivity of the assessments of the strength of a given factor that affects food consumption. In addition, our approach cannot assess the impacts of unknown factors. Because we did not provide interviewees with a predetermined list of factors, it is possible that many of them were unaware of some of the factors that other interviewees identified or failed to remember those factors during the interviews. However, this weakness of our research is also a strength, since the interviewees had a chance to identify factors that they considered to be important rather than being biased by our prior choices of factors. Future research should focus on the complete set of factors that we identified in the present study.
Our results are also influenced by various subjective biases. For example, residents tended to remember the impact of recent or extreme climate events, and this may have exaggerated the weight they assigned to such factors. This shortcoming could be mitigated by conducting surveys with a time lag in future research to see whether perceptions change.

Conclusions
In this study, we used primary data obtained by conducting interviews to support fuzzy cognitive mapping in three typical areas of the Xilin Gol Grassland. We identified the factors that residents believed had affected their food consumption along a north to south transect through the grassland based on household perceptions of the key factors and their relative strengths, and combined those findings with socioeconomic data to analyze the impacts of the key factors in three typical regions of the grassland. From north to south along the transect, perceptions of the factors that affected food consumption changed in terms of both the key factors and their perceived strengths. The number of livestock, household income, regional economic development, consumption habits, age, and infrastructure were perceived as the most important factors. However, the residents of pastoral areas (West Ujimqin and Zhenglan) focused more on ecological and economic factors, whereas residents of the agro-pastoral transitional zone (Taibus) placed less emphasis on ecological factors and were more concerned about personal and social factors. The key factors differed among the three regions, which suggests that local factors influenced the residents' perceptions and the resulting food consumption.
Our data on the factors that influence food consumption are preliminary, but future research to improve our understanding of these household perceptions will provide insights into the relationships between human activities and the grassland ecosystem, thereby providing improved guidance for policy development.