A Geospatial Framework of Food Demand Mapping

Abstract

Spatial mapping of food demand is essential for understanding and addressing disparities in food accessibility, which significantly impact public health and nutrition. This research presents an innovative geospatial framework designed to map food demand, integrating individual dietary behaviors with advanced spatial analysis techniques. This study analyzes the spatial distribution of eating habits across Lithuania using a geospatial approach. The methodology involves dividing Lithuania into 60,000 points and interpolating survey data with Shepard’s operator, which relies on a weighted average of values at data points. This flexible approach allows for adjusting the number of points based on spatial resolution and sample size, enhancing the reliability and applicability of the generated maps. The procedure includes generating a structured grid system, incorporating measurements into the grid, and applying Shepard’s operator for interpolation, resulting in precise representations of food demand. This framework provides a comprehensive understanding of dietary behaviors, informing targeted policy interventions to improve food accessibility and nutrition. Traditional food spatial mapping approaches are often limited to specific polygons and lack the flexibility to achieve high granular detail. By applying advanced interpolation techniques and ensuring respondent location data without breaching privacy concerns, this study creates high-resolution maps that accurately represent regional differences in eating habits. The methodology’s flexibility allows for adjustments in spatial resolution and sample size, enhancing the maps’ validity and applicability. This novel approach facilitates the creation of detailed food demand maps at any granular level, providing valuable insights for policymakers and stakeholders. These insights enable the development of targeted strategies to improve food accessibility and nutrition. Additionally, the obtained information can be used for computer simulations to further analyze and predict food demand scenarios. By leveraging spatial data integration, this study contributes to a deeper understanding of the complex dynamics of food demand, identifying critical areas such as food deserts and swamps, and paving the way for more effective public health interventions and policies aimed at achieving equitable food distribution and better nutritional outcomes.

1. Introduction

Understanding the dynamics of food demand is essential for addressing global challenges related to nutrition, health, and sustainability. As urban areas grow and populations increase, the demand for food intensifies, leading to significant challenges in ensuring food accessibility and equity. According to the United Nations, approximately 56% of the world’s population lived in urban areas in 2021, and this number is projected to rise to nearly 70% by 2050 [1,2]. Additionally, the global population is expected to reach 9.7 billion by 2050, further increasing the demand for food [3]. This intensifying demand requires targeted strategies to promote equitable access to nutritious food options, as urbanization continues to accelerate and populations expand. This situation is compounded by supply chain issues and economic imbalances, which contribute to the emergence of food deserts. Analyzing food demand patterns is crucial for crafting effective policies and interventions. This analysis informs agricultural production and supply chain management, and highlights the nutritional needs of diverse populations. Moreover, it sheds light on the environmental impact of food production, guiding efforts towards more sustainable practices. By exploring the factors that influence food demand, researchers and policymakers can better predict future needs, ensuring food security and promoting healthier eating habits across communities. The study by Liddy et al. (2023) discusses the limitations and opportunities in food mapping approaches, particularly in rapidly growing urban regions. It emphasizes the need for comprehensive frameworks that can capture the dynamic and multifaceted nature of food systems [4]. This aligns with the novelty of our work, which aims to provide a more granular and flexible spatial mapping of food demand.

1.1. Growing Urban Areas and the Emergence of Food Deserts

Comprehending the factors influencing food demand is vital for addressing international challenges in nutrition, health, and sustainable development. In the face of growing populations and changing dietary preferences, analyzing food demand patterns becomes crucial for crafting effective policies and interventions. This analysis not only informs agricultural production and supply chain management but also highlights the nutritional needs of diverse populations. Furthermore, this analysis reveals the environmental effects of food production, directing initiatives towards sustainability. Investigating the drivers of food demand enables researchers and policymakers to anticipate future requirements more effectively, securing food availability and fostering healthier dietary practices across different communities [5].

One critical aspect of this analysis is the understanding of “food deserts”, a concept that has its origins in urban planning in the UK but gained prominence in the United States. “Food deserts” are defined as areas characterized by poor access to healthy and affordable food, which may contribute to social and spatial disparities in diet and diet-related health outcomes [6]. Such areas are often characterized by reliance on fast-food chains and convenience stores, which offer limited healthy options. The absence of accessible grocery stores is exacerbated by a complex web of factors—low income, lack of transportation, and sometimes even cultural barriers—that contribute to poor dietary habits and subsequent health issues. In the U.S., the issue of food deserts has a distinct historical and socio-economic context. Urban decay and the exodus of supermarkets to more affluent suburbs have disproportionately impacted marginalized communities, especially African American and Hispanic neighborhoods. These systemic inequalities have led to higher rates of diet-related diseases like obesity and diabetes among residents of food deserts [5].

In response to these challenges, e-commerce platforms present a potential solution to the geographical limitations of food deserts. These platforms can offer residents access to a broader variety of food products than is typically available in their immediate neighborhoods. However, this potential is tempered by several challenges. Internet access, often a given in wealthier neighborhoods, may not be as readily available in marginalized areas. Additionally, the associated costs of online shopping, such as delivery fees and minimum purchase requirements, can be prohibitive for low-income households. Research adds another layer of understanding to this issue. A systematic review confirmed that disparities in food access in the U.S. were clearly tied to income and race, though similar evidence from other high-income countries was found to be limited [6]. One study took an innovative approach by creating a refined food desert index that leverages real-time data and machine learning, offering a more nuanced, dynamic picture of food environments [7]. Cooksey-Stowers, Schwartz, and Brownell expanded the discussion by introducing the concept of ‘food swamps’—areas where unhealthy food options significantly outnumber healthy ones. Their study found that food swamps were a more potent predictor of adult obesity rates, especially in areas with high income inequality and low resident mobility [8].

1.2. Evolving Consumer Purchasing Behavior

As consumer purchasing behavior continues to evolve, understanding these shifts becomes essential in addressing the broader issues of food accessibility and nutrition. The COVID-19 pandemic has led to significant changes in lifestyles, creating a larger gap between individuals who shifted towards healthier eating habits and those who increased their preference for high-calorie diets. This divergence highlights the need for targeted strategies to promote equitable access to nutritious food options. This change is particularly evident among teenagers, who now favor online communication with their friends, likely due to the increased reliance on virtual meetings during the pandemic for educational purposes and discussions [9,10]. Consequently, the frequent use of devices like smartphones has become a typical aspect of daily life for many adolescents, making owning such gadgets a standard practice. This alteration in lifestyle habits has led to an increase in sedentary behavior and a preference for high-calorie diets, which subsequently results in the prevalence of overweight conditions among individuals. It is important to understand that numerous other metabolic illnesses, such as type 2 diabetes (T2D), non-alcoholic fatty liver disease (NAFLD), cardiovascular diseases (CVDs), chronic kidney diseases (CKDs), and malignancies, are frequently linked to obesity [11,12]. Additionally, bad eating habits significantly impact society’s financial health and general well-being [13]. While previous research has often highlighted the negative impact of the pandemic on eating behaviors, showing an increase in the consumption of comfort foods and unhealthy snacking, other studies have demonstrated a shift towards healthier food choices among certain populations. For instance, Ammann et al. (2022) stated that “the possibility to work remotely led to healthier food choices, that is, a reported increase in vegetable consumption and decrease in sweet snack consumption” [14].

These shifts in dietary habits underscore the complexity of eating behaviors, which are not solely driven by the need to satisfy biological requirements for nutrients. Eating behaviors are also shaped by a complex interplay between individual preferences, interpersonal relationships, and the broader social environment [15]. Factors such as gender, age, and physical health are linked to food cravings, which in turn impact dietary habits [16]. Individual differences in eating behavior can be attributed to personality factors that influence eating habits and food choices [5]. Self-efficacy, defined as the belief individuals have in their capability to successfully carry out a specific task, plays a crucial role in influencing their eating habits [17]. Furthermore, physical activity is a behavior that promotes a healthier diet and is critical for controlling weight in various populations, including those who are normally weighted. This conduct affects physiological mechanisms like the regulation of appetite and psychological factors including self-efficacy and body image perception. These influences culminate in enhanced self-motivation, ultimately fostering better self-regulation in dietary practices [18,19]. Elevated physical activity levels could significantly enhance motivation regarding eating behaviors. Engaging in more physical activity might contribute to a higher likelihood of successful weight loss, partly due to its impact on dietary habits. This includes a more adaptable approach to food restrictions and a reduction in emotional eating patterns [20]. Conversely, reduced participation in physical activities is linked to eating behaviors driven by external motivations and correlates with a higher body mass index (BMI) [21].

In addition to individual factors, the social and community context plays a crucial role in shaping eating behaviors. Substantial evidence indicates that factors like place of residence, educational background, occupation, and income level significantly influence eating behavior, including the types of food chosen and the quantity consumed [22,23]. Understanding these multifaceted influences on consumer purchasing behavior is essential for developing effective strategies to promote healthier eating habits and address the broader issues of food accessibility and nutrition. As dietary habits evolve in response to various factors, it becomes increasingly important to consider both individual and community-level determinants in designing interventions aimed at improving public health outcomes.

1.3. Need for Spatial Representation of Food Demand

As consumer purchasing behavior continues to evolve, it is crucial to develop innovative approaches that can effectively identify and respond to these changes. To effectively identify consumer trends and provide actionable information for supply chain management, there is a critical need for spatial representation of food demand. Traditional methods of identifying food deserts often rely on census data that may be outdated or fail to capture real-time changes in food environments. Innovations, including the development of a refined food desert index, offer dynamic, real-time data that provide more nuanced and context-aware insights into the spatial aspects of food demand [7]. By leveraging advanced methodologies, researchers can gain a deeper understanding of environmental health disparities and design targeted interventions. Furthermore, ‘food swamps’, areas predominantly filled with high-calorie, low-nutrition food options, are emerging as potent predictors of adult obesity rates, even surpassing the effects of food deserts [8]. This highlights the need for comprehensive strategies that go beyond merely improving access to food and should also focus on the quality and type of food available in different regions, especially those with higher income inequality and lower resident mobility. Collectively, these studies emphasize the importance of a multifaceted approach to understanding the geographic and socioeconomic nuances of food demand [6,7,8].

In addition to identifying food deserts and swamps, research also highlights the spatial and socio-economic complexities affecting food demand and security. The ageing population in Germany significantly influences the food demand chain, impacting production, logistics, and retail due to changes in consumption patterns and the increased importance of catering to elder consumers’ specific needs [24]. For instance, studies exploring the links between food insecurity, socioeconomic factors, and BMI among young people in urban settings reveal the influence of low-income areas on food insecurity and higher BMI [25]. Similarly, a 2023 study delved into the impact of the COVID-19 pandemic on food insecurity in Los Angeles County, underscoring the role of geographical barriers such as restricted grocery store hours and lack of vehicle access in contributing to food insecurity [26]. Online grocery shopping offers both promise and pitfalls for healthier food purchases, potentially reducing unhealthy impulse buys while also presenting challenges such as consumers’ hesitance to purchase fresh produce online [27].

Despite the insights gained from these studies, there remains a gap in comprehensive studies that integrate demographic statistics with e-commerce insights to provide a spatial perspective on food demand. Traditional food demand forecasts often rely on aggregate data or expert opinions, missing the nuanced household-level variations that are critical for effective interventions [28]. Similarly, e-grocery simulations typically utilize random food demand patterns, ignoring the complex interplay of regional and environmental variables that can shape food demand [29].

Addressing the need for spatial representation of food demand is critical for developing effective and targeted strategies to improve food accessibility and nutrition. By integrating advanced methodologies and real-time data, researchers and policymakers can gain a comprehensive understanding of the geographic and socioeconomic factors influencing food demand. This approach will enable the more precise identification of areas in need, such as food deserts and food swamps, and inform the development of interventions that consider both individual and community-level determinants. Ultimately, such efforts will enhance public health outcomes and ensure equitable access to nutritious food for all populations.

1.4. Novel Food Demand Mapping Framework

Building on the necessity for spatial representation of food demand, this research introduces an innovative food demand mapping framework designed to overcome the limitations of current approaches. The framework consists of detailed food surveys, population sampling, and spatial interpolation techniques to create accurate and high-resolution maps of food demand. This innovative approach aims to bridge the gap between micro-level dietary choices and macro-level spatial analytics, enabling a comprehensive understanding of food demand. By integrating surveys on eating habits with advanced spatial mapping techniques, the framework seeks to identify and address disparities in food access, ultimately guiding more targeted and effective policy interventions.

To further elaborate, our research methodology advances the discourse on food demand mapping by integrating and refining elements from existing methodologies, focusing on granular-level data analysis. The work of Wang et al. (2015) employs spatial analysis to understand food provision across China, laying a foundational understanding of agro-ecosystems at a broad scale. They found that South and Southeast China have higher ecosystem food provision potential (EFPP) than the North, but the North, especially Northeast China, has a higher conversion ratio of EFPP (CRFP). This suggests the need to protect and explore food potential in the South while acknowledging the nearing limits of food provision in the North due to high-intensity cropping [29]. Similarly, Hood et al. (2020) utilize demographic and geographic data from a vast survey to analyze e-commerce in grocery shopping, highlighting the importance of integrating spatial data for understanding consumption patterns. Their findings show that home delivery is the dominant e-commerce channel, particularly among females, individuals aged 25–44, and affluent groups. Geographic analysis reveals higher adoption rates in rural areas, suggesting that logistics and infrastructure investments should consider local consumer behavior patterns [30]. Our approach parallels the comprehensive use of household data by Wichern et al. (2018) to reveal food security patterns in Uganda, emphasizing the value of granular data. Wichern et al. found significant local variation in food security indicators, making large-scale predictions challenging. Their study demonstrates that environmental variables alone can provide considerable predictive power and that local differences in food availability and livelihood activities are crucial for effective interventions [28]. Furthermore, Buscemi et al. (2023) explore the spatial interconnections between food insecurity, BMI, and socio-economic factors, underscoring the significance of detailed spatial analysis in identifying health disparities. They discovered strong links between food insecurity and low income, with higher BMI associated with lower income and fewer grocery stores, and significant geographic disparities in these factors across Chicago [25]. By integrating these diverse methodologies and findings, our research aims to provide a nuanced understanding of food demand mapping and its implications for policy and intervention strategies.

What sets our approach apart is the ability to transcend these methodologies by spatially interpolating survey results at a more refined level. This allows for unprecedented flexibility in analyzing food demand distribution. This enables a nuanced understanding that captures both macro- and micro-level dietary preferences and accessibility issues, facilitating precise identification of food deserts and swamps. By doing so, we contribute a unique lens through which food demand can be mapped, blending the rigorous spatial analyses seen in previous studies with an innovative approach to granularity.

This investigation distinguishes itself through the innovative integration of geographic interpolation techniques within the domain of medical research, thereby pioneering a novel methodological framework for the spatial delineation of regions beset by nutritional challenges. Building upon the foundational work of Wang et al. [29], which employs spatial analysis to understand food provision across China, our study extends the application of spatial methodologies to the nutritional domain. Similarly, drawing on the insights of Hood et al. [30], who utilize demographic and geographic data to analyze e-commerce in grocery shopping, this study underscores the importance of integrating spatial data for comprehending dietary patterns. Our distinctive approach thus enables the precise identification of areas most in need of nutritional interventions, facilitating targeted and efficacious health-promoting strategies.

Our study, functioning as a foundational pilot investigation, particularly emphasizes the methodological innovation it introduces, harnessing the capabilities of rapidly growing fields like spatial statistics and geoinformatics to enable the creation of thematic maps. These maps offer a means to determine the variability and accuracy of measurements, presenting a comprehensive overview of the entire geographical area. Typically, measurements are conducted in natural studies to elucidate field characteristics. In the specified study [31], thematic maps are meticulously constructed to discern fluctuations in soil properties, vegetation, and yield based on samples gathered throughout the crop cycle. This precision in agriculture, plant cultivation, or the modeling of heavy metals facilitates the judicious selection of planting locations for strategic planning and risk control. Various interpolation methods, such as natural-neighbor interpolation, inverse functions of distance, least-squares polynomials, and kriging, are employed for modeling purposes [31].

According to researchers [32], the conventional statistical approach typically requires adherence to specific assumptions, such as the independence of observations, the exact or approximate normality of observations, and the necessity for extensive and repetitive sampling. For instance, autocorrelation, a crucial aspect highlighted by the kriging method, relies on a normal distribution, making it suitable for modeling natural processes. However, not all conducted studies reveal a normal distribution, limiting the research and its expansion. In such cases, this work opts for Shepard’s operator or Inverse Distance Weighting (IDW), a widely acknowledged method valuable in modeling results from various perspectives [32,33]. Shepard’s operator is an interpolation technique that assigns values to unknown points based on the weighted average of known points, where the weight decreases with distance. Similarly, IDW is a spatial interpolation method that estimates cell values by averaging the values of nearby points, giving more influence to closer points. Both methods are essential in spatial analytics for creating smooth and accurate surfaces from scattered data. In study [34], Shepard’s modeling was utilized in the generation of noise maps in the city of Isparta, Turkey, which were analyzed with consideration given to grid resolution. This approach integrates seamlessly with the framework of spatial data interpolation, evaluating a region’s average pollution level, wherein the selection of the appropriate exponent value for each real-world scenario has been methodically determined [35]. Preliminary modeling across various pollution scenarios has demonstrated that the selection of the distance exponent must be contingent upon the specific issue at hand, rather than being based on an arbitrary decision [35]. For this reason, we selected an exponent that enables the appropriate visual representation of maps.

The novelty of this research lies in its presentation of dietary data through a mapped format, utilizing the Eating Habits Scale. This methodological innovation parallels the comprehensive use of household data by Wichern et al. [28] to reveal food security patterns in Uganda, and is further informed by the work of Buscemi et al. [36], who explore the spatial interconnections between food insecurity, BMI, and socio-economic factors. The significance of this study extends beyond its methodological contributions; it presents a pivotal opportunity for the establishment of cross-national collaborations aimed at assessing dietary patterns across diverse geographical contexts. Such international research endeavors, informed by the granular-level data analysis advanced in our methodology, hold the potential to unveil region-specific dietary behaviors, thereby informing the development of tailored nutritional interventions. Through this collaborative and targeted approach, the study not only advances the scientific understanding of nutritional epidemiology but also contributes to the global effort in improving public health outcomes related to diet and nutrition.

However, it is important to note the primary limitation of this study: its reliance on a small sample size. This was a deliberate choice given its nature as a pilot study aimed at testing the proposed food demand mapping framework. Future research will aim to expand the sample size to enhance the robustness and generalizability of the findings. Looking ahead, the research intends to develop an agent-based model of the food supply chain, utilizing the insights gained from the mapping exercise. To circumvent the challenges associated with conducting extensive surveys, we plan to integrate satellite imagery as a novel approach for predicting food demand at a granular level. This methodology promises to streamline the data collection process, offering a more efficient and scalable means of understanding food demand dynamics without the logistical complexities of survey distribution and collection.

The primary goal of this research is to develop a comprehensive framework for mapping food demand, integrating individual dietary behaviors with spatial data to address disparities in food access. To validate the proposed framework a case example of Lithuania will be conducted. The specific objectives of this study are:

• Create a robust methodology for mapping food demand based on survey data that capture both individual and regional consumption patterns.
• Implement a pilot survey in Lithuania to gather preliminary data on individual eating habits and preferences, laying the groundwork for more comprehensive future studies.
• Employ advanced spatial analysis techniques to generate visual maps representing food demand across different regions, with a focus on identifying and highlighting areas of concern such as food deserts and food swamps, thereby validating the proposed framework.

2. Materials and Methods

2.1. Food Demand Mapping Framework

The food demand mapping framework is designed to synthesize individual dietary preferences with geospatial data, offering a multidimensional view of food accessibility and consumption patterns (see Figure 1). This innovative approach aims to bridge the gap between micro-level dietary choices and macro-level spatial analytics, enabling a comprehensive understanding of food demand. By integrating detailed surveys on eating habits with advanced spatial mapping techniques, the framework seeks to identify and address disparities in food access, ultimately guiding more targeted and effective policy interventions. This introduction to the framework sets the stage for a nuanced exploration of the complex interplay between personal food choices and geographic food availability.

2.2. Ethics Approval

The anonymous internet survey conducted in Lithuania did not require ethical approval according to local laws, as it did not qualify as a biomedical study. During the survey, the researchers were not privy to any personally identifiable information, ensuring that all participant responses were kept confidential and anonymous.

2.3. Study Design

This research is a cross-sectional study that utilizes convenience sampling. For this study, one survey was created and administered using Apklausa.lt (https://apklausa.lt/).

2.4. Inclusion and Exclusion Criteria

This research included individuals living in Lithuania with internet access. Participants were not personally contacted or offered any incentives to participate in the study. In this study, no responses were excluded due to incorrectly completed surveys. The survey participants were individuals with internet access, which is notably extensive in Lithuania, a country recognized for its exceptional internet availability and speed within Europe [37]. Consequently, a substantial portion of the population was able to participate. The survey remained open for one month and was evenly distributed, with 100 surveys allocated per region. It is important to acknowledge that individuals lacking internet access were inherently excluded from this study, as their participation was not feasible.

2.5. Instruments

The study was structured into separate sections for a detailed examination of dietary habits and the adoption of online grocery shopping across different demographics. The first part, Demographic Profiling, collected basic demographic information such as age, gender, occupation, education, location, income, and BMI for contextual analysis. The second part, E-Grocery Shopping Preferences, delved into online grocery shopping habits, including preferences for delivery times, responses to delays, expectations for perishable items, shopping frequency, and handling of missed deliveries. The Healthy Eating Assessment questionnaire, developed by the Government of Northwest Territories, was utilized to assess eating habits. This approach involved participants rating their consumption of various food categories, such as fast/fried foods, fruits, vegetables, sugary beverages, snack chips, sweet foods, dairy products, and meat/fish/beans. The assessment utilizes a scoring system, where respondents circle their answers for each question, and the total score is used to determine their health benefit zone. This standardized tool consists of 10 questions focused on dietary behaviors. The outcomes of the questionnaire are compiled into a scoring system, where a total score of 10–19 indicates a need for improvement, 20–29 is considered fair, 30–39 is good, and a score between 40 and 50 reflects excellent eating habits [26].

2.6. Statistical Analysis

To provide a geographical perspective of the results, a map was developed. This map illustrates the dietary habits across various neighborhoods, highlighting areas with exemplary dietary practices and pinpointing those where improvements in dietary habits are necessary. The development of this map was executed using R 4.2.2. software on a computer with the main processor parameters Intel (R) CORE (TM) i7-2600 CPU@3.40GHZ and memory 8 GB ensuring accuracy and clear visualization. The use of these analytical tools was instrumental in offering a comprehensive view of the eating habits throughout the studied regions, yielding valuable insights for research and potential health interventions.

2.7. Integration with Population Data

During the initial data preprocessing phase, participants’ ages were categorized into predefined brackets. These age brackets, along with other demographic information, were meticulously aligned with the Lithuania National Statistical Department’s population survey data. It was noted that the national statistical department did not cover some elderly municipalities in their 1 × 1 km grid data. For respondents from these areas, or those who did not specify their elderly municipality, points were generated based on the broader municipality region. The 1 × 1 km grid from the Lithuanian Statistical Department’s population census served as a foundational dataset [38]. This grid, which detailed demographic distributions like age and gender, allowed for a seamless fusion of the survey data, thereby ensuring that the spatial representation mirrored the demographic dispersion of the population accurately. To represent respondents spatially without violating privacy, an innovative approach to generating respondent coordinates was devised. Initially, the grid data were filtered based on each respondent’s age and gender. Subsequently, relevant boundaries were identified depending on the provided municipality and elderly municipality names. Within these boundaries, grids were selected randomly. The selection’s randomness was weighted primarily by the age demographic to ensure a representative selection. However, the methodology also boasts versatility by allowing for the incorporation of other demographic statistics as additional weights, if required. For each selected grid, a random point was generated to symbolize the respondent.

2.8. Food Map Generation

In this study, the spatial distribution of eating habits across Lithuania will be analyzed using a geospatial approach. The methodology involves dividing Lithuania into 60,000 points and interpolating survey data. The number of points can be adjusted based on the required spatial resolution and the sample size, which might impact the validity of the map. This flexibility allows for more precise and accurate representations of food demand, catering to specific research needs and ensuring that the spatial analysis aligns with the available data. Adjusting the number of points helps in balancing the granularity of the analysis with the practical limitations of survey data, ultimately enhancing the reliability and applicability of the generated maps.

Our main aim is to analyze the spatial distribution of eating habits across Lithuania, employing a geospatial approach. Our methodology involves dividing the country into 60,000 points and utilizing spatial interpolation with survey data. As already discussed in the previous sections, a well-established deterministic approach for scattered data interpolation is Shepard’s operator. This operator relies on a weighted average of values at data points and, owing to its simplicity, finds frequent application in interpolating natural processes like groundwater depth or pollution, as documented in [33].

Consider dataset $X=(x_1,x_2,\dots ,x_k)$, where $x_i\in {\mathbb{R}}^p$ for $1\le i\le k$ and $p\ge 1$. The corresponding response values are given by $Y=(y_1,y_2,\dots ,y_k)$ where $y_i\in R^m$, $p$ and $m$ denotes the number of features, and $k$ represents the sample size. Let us define a linear extrapolator ${y:R}^p\to R^m$ as follows:

$$y\left(x\right)=Y^{T}u\left(x\right)$$

where

$$u(x)=\left\{\begin{array}{cc}{\displaystyle \frac{w\left(x\right)}{1^{T}w\left(x\right)}},& if \left|x_i-x\right|\ne 0 \forall i\\ y_i , & if \left|x_i-x\right|=0\end{array}\right.$$

where $\mathbf{1}$ is a $k$—dimensional vector of ones, and the weights are determined by:

$$w\left(x\right)=\left({\displaystyle \frac{1}{{\left|x_1-x\right|}^{\delta }}},{\displaystyle \frac{1}{{\left|x_2-x\right|}^{\delta }}},\dots ,{\displaystyle \frac{1}{{\left|x_k-x\right|}^{\delta }}}\right).$$

with $0\le \delta \le 1$. Note that the Shepard’s extrapolator satisfies the condition ${\mathbf{1}}^{\mathit{T}}u\left(x\right)=1$.

2.9. Map Creation Procedure

The following procedure outlines the steps for creating detailed maps by integrating both spatial and data-driven elements:

• Generate Grid:

A structured grid system utilizing the shape file of the specific country of interest is created. This grid establishes a spatial framework crucial for subsequent stages of data interpolation. For this, $k=70; 112$ observations were applied for model training. The prediction grid had 60 000 rows and 2 columns, with 0.21 × 0.21 km and 1.04 × 1.04 km spacing covering the study area that is Lithuania and Vilnius region.

• Input into Shepard’s Interpolator:

The measurements have been incorporated into the pre-existing grid. Concurrently, the Shepard’s exponent is presently under careful consideration for selection.

• Interpolation to Grid Points:

Shepard’s operator to interpolate values across the entire grid is applied. This step results in a representation of geographic data.

3. Results

The results of this study offer a detailed examination of the spatial distribution of eating habits, providing a comprehensive understanding of how dietary behaviors vary across different regions. By employing advanced geospatial analysis techniques, we have been able to map food demand with high precision, highlighting significant regional disparities. This chapter presents the findings from our spatial mapping efforts and discusses the implications of these patterns for public health and policy-making.

3.1. Individual Behavior

The survey results focusing on e-groceries have been published in another journal [39], as the small sample size (n = 120) in this pilot study does not allow for a comprehensive generalization of food preferences. Although these preliminary findings are limited, they provide valuable insights into the emerging trends and behaviors associated with online grocery shopping. Participants in the pilot study demonstrated a varied range of e-grocery usage patterns, highlighting the potential for significant shifts in consumer behavior as digital marketplaces expand. In future studies, expanding the survey to include a larger and more diverse sample size will be crucial. This will enable the development of more generalized insights that can accurately reflect broader population trends and preferences. A larger dataset will allow for more robust statistical analyses, facilitating the identification of key factors influencing e-grocery adoption and the specific food preferences of different demographic groups. Additionally, it will enhance the reliability of the data, enabling researchers to draw more definitive conclusions about the role of e-groceries in shaping modern dietary habits. By ensuring the Section 2 and Section 3 focus on spatial mapping and population data integration, we align the publication with the intended separation of survey raw data interpretation. This adjusted content should provide a clear and focused narrative for your research.

3.2. Food Regional Mapping

The Section 3 unveils the spatial distribution of eating habits across Lithuania, delineating areas of nutritional excellence and concern through a color-coded mapping approach. This analysis, grounded in a detailed examination of the geographical nuances in dietary patterns, identifies significant regional disparities. With northern Lithuania showcasing healthier eating habits and southern regions indicating areas of nutritional deficit, the study provides a comprehensive overview of the country’s dietary landscape. The mapping further reveals a gradient of dietary habits within Vilnius, suggesting socio-economic and accessibility factors at play. This foundational understanding sets the stage for targeted interventions and policy-making, aiming to bridge the nutritional divide across regions (Figure 2).

Utilizing the Shepard interpolation method, the primary outcomes are presented in

Table 1. Actual data and interpolation results.

	Min	Max	Median	Mean
Actual data (Lithuania)	21.00	42.00	33.00	32.84
Actual data (Vilnius)	21.00	42.00	32.00	32.62
Results (Lithuania)	21.00	42.00	32.88	33.02
Results (Vilnius)	21.00	42.00	32.66	32.75

. Analysis of the mean and median values reveals that the predictive outcomes are in consistent agreement with the sample of measured data points, evidencing a notable accuracy level. In the context of Lithuania as a whole, the observed data yielded a mean of 32.84 and a median of 33.00. The predictive outcomes demonstrated a mean of 33.02 and a median of 32.88, indicating minimal discrepancies. Within the specific locale of Vilnius county, the observed measurements indicated a mean of 33.02 and a median of 32.88, while the predictive results showed a slightly lower mean of 32.75 and a median of 32.66. These differences are slight. Moreover, the interpolation method successfully identified the minimum and maximum values for both the expansive territory of Lithuania and the considerably smaller region of Vilnius county. This accuracy in capturing the extremities of the data further validates the efficacy of the Shepard interpolation method in producing reliable predictive insights. Based on the map provided in the image of Figure 2 we can discern the spatial distribution of eating habit scores across the geography of Lithuania. The color spectrum on the right indicates the total score, with lower scores in blue signifying poorer eating habits and higher scores in red to orange representing healthier eating habits (Figure 2).

In the northern part of Lithuania, the map shows a predominantly orange hue, suggesting that the populations here have better eating habits with higher total scores. Conversely, the southern part of Lithuania exhibits pockets of blue, particularly in the southeast, indicating regions where eating habits are not as favorable, reflected by lower total scores (Figure 2).

Moving to the western part of the country, we observe a blend of colors, but with a notable presence of yellow to orange shades, indicating moderate to good eating habits. The eastern part, similar to the southern region, displays a mixture of hues with instances of both blue and orange, suggesting a variable distribution of eating habits, where some areas have healthier practices than others. However, in the south-western part of Lithuania, there is a clear blue area indicating poor nutritional intake, and it is in this region that invasive methods are needed to enlighten the people living in that region (Figure 2).

To summarize, the best eating habits, as indicated by higher scores in Table 1 and maximum values and depicted in red to orange hues, are predominantly found in the northern and western parts of Lithuania. The worst eating habits, denoted by the lower scores in Table 1 and minimum values in blue, are primarily observed in the southern part, with particular emphasis on the southwest regions of the country. These interpretations are based on the color-coded map provided in Figure 2 and Figure 3, which are common methods in scientific studies for visualizing and interpreting geospatial data related to health behaviors and outcomes.

In the context of Vilnius city, the central areas exhibit a concentration of blue, which suggests that the central localities have lower eating habit scores as presented in Table 1, specifically recorded at 21.00, indicative of less favorable dietary practices. These areas of lower scores marked by both average and median scores (Table 1) are enveloped by regions of yellow, transitioning to orange, implying a gradient of improvement in eating habits as one moves outward from the city center. The peripheral areas of Vilnius, particularly those in the outlying regions, display an orange hue, which correlates with higher eating habit scores where the maximum is 42.00. This pattern may be indicative of better dietary practices or greater adherence to nutritional guidelines in these zones (Figure 3).

To provide a more nuanced interpretation, the gradation from blue in the city center to orange in the outskirts may reflect socio-economic, cultural, or accessibility factors influencing dietary choices in capital city. These findings underscore the heterogeneity of eating habits within the city, suggesting a spatial dependency that merits further investigation to elucidate the underlying determinants (Figure 3).

In the context of the spatial distribution of dietary habits within Lithuania, it is noteworthy to elucidate the intra-regional variances observed within the Vilnius area. The heatmap analysis delineates a dichotomy within the city: the south-eastern quadrant is characterized by an aggregation of red to orange hues, indicative of commendable eating practices among the inhabitants. This area’s higher scores are reflective of dietary habits that are aligned with nutritional guidelines, thereby representing public health success within the urban landscape (Figure 3).

In stark contrast, the north-eastern section of Vilnius presents a significantly different profile, with blue hues predominating. This signifies a prevalence of suboptimal eating habits that necessitate public health attention and intervention. The divergence between the south-eastern and north-eastern sectors of the city is pronounced and suggests a geographical health disparity within the urban confines (Figure 3).

Moreover, the western part of Vilnius is also marked by favorable eating habits, as evidenced by the presence of red to orange shades on the heatmap. This pattern denotes a consistent adherence to dietary recommendations, which is indicative of a well-established culture of health consciousness in these locales (Figure 3).

The north and north-eastern regions of Lithuania, extending beyond the metropolitan boundaries of Vilnius, emerge as areas of concern. The observed blue regions in these parts signal the urgent need for targeted nutritional behavioral interventions. These interventions are imperative to address poorer eating habits and to promote a more wholesome dietary culture, with the ultimate goal of enhancing the overall health and well-being of the population (Figure 3).

To streamline food system enhancements across various sectors, it is crucial for policymakers to lead the charge by crafting policies that promote the establishment of grocery stores within food desert regions, as pinpointed by advanced food demand mapping. This initiative could be further bolstered by incentivizing e-grocery services to extend their services to economically disadvantaged neighborhoods, alongside investments in infrastructure and internet accessibility to bolster the efficacy of these digital solutions. Farmers, on the other hand, stand to gain from tailoring their crop production to align with the granular insights unearthed by food demand mapping, focusing on cultivating crops that cater to the specific dietary needs and preferences of diverse regions. This precision in farming can be augmented through partnerships with local governments and businesses to forge direct-to-consumer avenues, enhancing the direct flow of nutritious food. In parallel, retail and e-grocery sectors are encouraged to leverage food demand maps to fine-tune their inventory and marketing efforts, ensuring that their product offerings resonate with the unique needs of their demographic and regional customer base. By adopting flexible pricing and delivery frameworks, these platforms can make healthy food alternatives more accessible to a broader audience, potentially offsetting the challenges posed by food deserts. Warehouse and transportation providers play a pivotal role in this ecosystem by optimizing their logistics and distribution strategies based on spatial food demand analyses. A focus on reducing delivery times and costs in areas identified as high-demand can significantly improve food access, underpinned by investments in technology to anticipate demand fluctuations and adapt logistical operations accordingly. This concerted effort across sectors is key to mitigating food accessibility issues and fostering healthier communities.

4. Discussion

In this research, we present an innovative approach to analyzing dietary intake variations among various socio-demographic cohorts. Utilizing this methodology, we constructed a geographic map of eating habits, enhancing the initial dataset and enabling the prediction of dietary behaviors by region. This methodological innovation, tested within the Lithuanian context, is adaptable to international models, facilitating the identification of areas where investment in nutritional education and resources is most warranted.

The spatial analysis of eating habits across Lithuania, utilizing geospatial interpolation methods such as Shepard’s operator, provides novel insights into the geographical distribution of dietary behaviors. This approach, which has been effectively applied in environmental sciences, offers a valuable tool for public health research by identifying regions with poor dietary habits that may benefit from targeted nutritional interventions [40].

Our research methodology advances the discourse on food demand mapping by integrating and refining elements from existing methodologies, focusing on granular-level data analysis. The work of Wang et al. [29], which employs spatial analysis to understand food provision across China, lays a foundational understanding of agro-ecosystems at a broad scale. Similarly, Hood et al. [30] utilize demographic and geographic data from a vast survey to analyze e-commerce in grocery shopping, highlighting the importance of integrating spatial data for understanding consumption patterns. Our approach parallels the comprehensive use of household data by Wichern et al. [28] to reveal food security patterns in Uganda, emphasizing the value of granular data. Furthermore, Buscemi et al. [36] explore the spatial interconnections between food insecurity, BMI, and socio-economic factors, underscoring the significance of detailed spatial analysis in identifying health disparities.

This innovative geospatial framework for mapping food demand addresses a significant research gap where current food spatial mapping approaches are often restricted to specific polygons and fail to capture finer granular details. Our study proposes an approach that allows for high-resolution mapping, offering a dynamic and context-aware representation of food environments, as emphasized in recent studies that highlight the importance of such advancements for more effective decision-making in food systems [4].

Our novelty lies in transcending these methodologies by spatially interpolating survey results at a more refined level, allowing for unprecedented flexibility in analyzing food demand distribution [41]. This enables a nuanced understanding that captures both macro- and micro-level dietary preferences and accessibility issues, facilitating precise identification of food deserts and swamps [42]. By doing so, we contribute a unique lens through which food demand can be mapped, blending the rigorous spatial analyses seen in previous studies with an innovative approach to granularity [43].

The primary limitation of this study is its reliance on a small sample size, which was a deliberate choice given its nature as a pilot study aimed at testing the proposed food demand mapping framework. The primary goal was to propose a framework for food demand mapping and not to derive precise insights, thus this does not reduce the primary contribution of the publication. The small sample size can introduce biases, as it may not be representative of the broader population, potentially skewing the findings and limiting their generalizability. The limited sample might also overlook variations in food demand patterns that would be evident in a larger, more diverse sample. Future research will aim to expand the sample size to enhance the robustness and generalizability of the findings. Increasing the sample size will help to mitigate the biases associated with small sample studies, providing a more accurate reflection of the population’s food demand. Additionally, a larger sample will allow for more detailed sub-group analyses, improving the understanding of different factors influencing food demand. Looking ahead, the research intends to develop an agent-based model of the food supply chain, utilizing the insights gained from the mapping exercise. To circumvent the challenges associated with conducting extensive surveys, we plan to integrate satellite imagery as a novel approach for predicting food demand at a granular level. This methodology promises to streamline the data collection process, offering more efficient and scalable means of understanding food demand dynamics without the logistical complexities of survey distribution and collection. However, it is important to note that, while satellite imagery can provide valuable data, it may also introduce its own set of biases, such as the potential misinterpretation of imagery data and the limitations of detecting certain food demand indicators from space. Future studies should consider these potential biases and develop strategies to address them, ensuring the accuracy and reliability of the findings. Another promising direction for future research is to integrate socio-economic indicators with food preferences obtained from surveys. By combining this information with census data, we could estimate actual food demand in precise quantities, rather than simply identifying trends. This approach would allow for a more comprehensive and accurate assessment of food demand, enabling policymakers and stakeholders to make more informed decisions regarding food security and resource allocation.

5. Conclusions

This research introduces a novel framework for mapping food demand by integrating individual dietary behaviors with advanced spatial data analysis techniques. Through the application of the Shepard interpolation method, we have successfully generated high-resolution maps that highlight regional disparities in eating habits across Lithuania. This granular approach allows for the precise identification of areas, such as food deserts and food swamps, which are in urgent need of nutritional interventions.

Our findings reveal significant regional variations in dietary practices, with the northern and western parts of Lithuania exhibiting healthier eating habits compared to the southern regions. In Vilnius, the spatial distribution of dietary habits shows a gradient from less favorable practices in the city center to healthier behaviors in the outskirts, reflecting socio-economic and accessibility factors. These insights are crucial for informing targeted policy interventions aimed at improving public health outcomes and ensuring equitable access to nutritious food.

Based on our findings, we propose the following specific policy recommendations: Firstly, policymakers should incentivize businesses to open grocery stores in underserved areas through subsidies, tax breaks, or grants. Secondly, e-grocery platforms should be encouraged to serve economically disadvantaged neighborhoods by subsidizing delivery costs and improving digital infrastructure. Thirdly, support should be provided to farmers to align crop production with the dietary needs identified in food demand maps through local government and business partnerships. Additionally, implementing community-based nutrition education programs tailored to specific cultural and socio-economic contexts is crucial. Enhancing public transportation to improve access to grocery stores in rural and low-income urban areas is also recommended. Retail and e-grocery sectors should offer discounts or free delivery for low-income households and accommodate working families’ schedules. Lastly, investing in local food hubs to aggregate and distribute produce from small farmers within the community will be beneficial.

The innovative methodological framework presented in this study not only advances the field of food demand mapping but also provides a robust tool for policymakers to address nutritional challenges. By leveraging detailed survey data and sophisticated spatial analysis, our approach transcends traditional methods, offering a dynamic and context-aware representation of food environments.

However, the study’s reliance on a small sample size, characteristic of a pilot study, poses limitations on the generalizability of the findings. Future research will aim to expand the sample size and incorporate additional data sources, such as satellite imagery and socio-economic indicators, to enhance the accuracy and reliability of food demand predictions. This expanded approach promises to provide a more comprehensive understanding of food demand dynamics, facilitating more effective decision-making in food systems and contributing to global efforts in achieving food security and promoting healthier eating habits.

In summary, this research underscores the importance of integrating micro-level dietary choices with macro-level spatial analytics to address food accessibility and nutrition challenges. The developed framework sets a foundation for future studies to build upon, offering a path forward for more precise and impactful nutritional intervention. In conclusion, this research represents a significant step forward in the application of geospatial analysis to the study of food demand and security. By providing a comprehensive and detailed view of dietary habits across Lithuania, it contributes valuable insights into the factors that influence food accessibility and consumption patterns. In forthcoming investigations, our intention is to broaden the regional survey and employ a novel fractional kriging method, which remains uninfluenced by a normal distribution and unveils the fractal interconnection and chaotic nature of the data. Ultimately, the goal is to leverage this knowledge to inform more effective policy decisions and interventions that ensure equitable access to nutritious food for all communities.