Comprehensive nutritional, environmental, and economic assessment for 3,302 recipes in China Authors: Yin Long, Xiaodan Shen, Ziyi Zhou, Niu Niu, Liqiao Huang, Fanrong Chen, Tianyao Shen, Yuyan Huang, Lingkang Xu, Shiqi Tan, Yufan Liu, Zhenyiran Tian, Yangjiashen Li, Kexin Liu, Yujie Wang, Meng Yu, Junyuan Huang, Jingyi Jiang, Minghao Zhuang, Xiangbo Xu, Qiran Zhao, Yoshikuni Yoshida Published: 2026 ABSTRACT AbstractDietary choices are fundamental to both human health and planetary sustainability, yet most evaluations still operate at the level of ingredients or food groups. This perspective neglects the cultural and culinary contexts that shape consumption and fails to capture the transformations induced by cooking methods. Such gaps limit our ability to assess trade-offs between nutrition, environmental sustainability, and economic affordability in ways that reflect real-world diets. China, with its rich and regionally diverse food system, provides a critical setting for advancing dish-level sustainability assessments. Our previous work established systematic profiles for Sichuan and Cantonese cuisines, but comprehensive coverage of other major culinary traditions has been lacking. The absence of such data constrains evidence-based strategies for culturally relevant dietary transitions. Here we fill this gap by compiling and standardizing a dataset of 3,302 recipes across six additional cuisines-Shandong, Jiangsu, Zhejiang, Fujian, Hunan, and Anhui. Each record includes detailed nutrient profiles, life-cycle environmental footprints, and market-based cost estimates, offering a robust foundation for comparative, dish-level sustainability analysis.Subject terms: Environmental impact, Interdisciplinary studies Introduction The global food system is a central determinant of both human well-being and planetary health1–3. Current modes of food production and consumption, however, are primary drivers of environmental degradation4. Agricultural expansion is linked to nearly 90% of global deforestation, with farm systems now occupying approximately 50% of the world’s habitable land5,6. The sector is also responsible for 70% of global freshwater withdrawals and is a leading source of greenhouse gas emissions6. In 2022 alone, agrifood systems emitted 16 billion tonnes of carbon dioxide equivalent (Gt CO2eq)7. Concurrently, dietary patterns pose a severe threat to public health, creating an urgent need for high-resolution data to monitor these risks. Poor diets have become the leading risk factor for the global burden of chronic non-communicable diseases, including obesity, cardiovascular conditions, and diabetes8,9. In 2021, dietary risks were attributable to over 13 million deaths, underscoring their dominant position in the human health risk landscape10. The EAT-Lancet Commission’s ‘Planetary Health Diet’, proposed in 2019, offers a normative framework for aligning human nutrition with planetary boundaries8. It has become a key benchmark for dietary sustainability, particularly in high-income nations11,12. However, its universal application is fundamentally constrained by two factors: economic affordability and cultural compatibility13–15. The high cost of this diet renders it inaccessible for an estimated 2.8 billion people, while its food-group-based recommendations often fail to translate to complex culinary traditions16. Effective dietary transitions demand place-based solutions grounded in local food systems and sociocultural contexts17. This requires shifting the data collection and analytical unit from abstract food groups to the culinary dish. The dish, not the ingredient, is the fundamental unit of consumption, embodying a complex interplay of nutritional composition, cooking methods, and cultural preferences3,18. Establishing a standardized database at this level is therefore essential for developing interventions that are both effective and culturally acceptable. China’s food system provides a critical model for developing place-based dietary data resources, owing to its unparalleled regional diversity. This diversity is embodied in its ‘Eight Great Cuisines’, a complex system of culinary traditions shaped by distinct geographies and local resources, resulting in significant heterogeneity in ingredients and preparation methods15,19,20. Our previous research provided a systematic assessment of Sichuan and Cantonese cuisines21. The present study expands the data coverage to six other major traditions: Shandong, Jiangsu, Zhejiang, Fujian, Hunan, and Anhui. This combination was strategically selected to represent a geographically continuous and culturally interlinked culinary spectrum, spanning northern to southern and coastal to inland China. The breadth of this spectrum is exemplified by the contrast between Shandong cuisine, the foundational northern style known for its emphasis on technique and rich stocks15,22, and Jiangsu cuisine, originating from the Jiangnan region, which is renowned for its meticulous knife skills and techniques that enhance the natural, delicate flavors of fresh river produce23,24. While these regional cuisines represent a rich cultural asset, they also present a sustainability trilemma when evaluated against contemporary health, environmental, and economic standards. Nutritional risks arise from specific culinary traditions, such as the high intakes of sodium and fat common in some northern and inland cuisines like Shandong, Anhui, and Hunan, or the refined use of sugar and starches in styles like Jiangsu and Zhejiang, both of which are linked to increased risks of cardiovascular and metabolic diseases25. The environmental footprints of these cuisines also vary substantially, driven by core differences in ingredient composition26; northern diets centered on livestock and grains thus exert a considerably greater pressure on land, water, and climate than the aquatic- and vegetable-based diets of the southeastern coast27. Furthermore, economic accessibility is often limited by the high material and labor costs of dishes that require rare ingredients or complex artisanship, such as the traditional banquet preparations found in Jiangsu and Fujian cuisines20,28. These inherent trade-offs between cultural heritage, health outcomes, environmental impact, and affordability necessitate a quantitative framework to assess their complex interdependencies. Assessing sustainability at the culinary dish-level is increasingly recognized as critical for developing effective, demand-side dietary interventions3,29. Pioneering research has validated this approach. For example, frameworks such as Germany’s RISE Institute30, Sweden’s LCA Food Database31, and the US Menus of Change initiative32, alongside localized studies in Mexico, India, and Japan, have begun to quantify the environmental footprints of specific cuisines33–35. However, existing research exhibits critical limitations that constrain its applicability. The heterogeneity of regional cuisines, in terms of cooking techniques, ingredient selection, and dietary preferences, is a primary determinant of a final dish’s nutritional and environmental profile36,37. Given that the fundamental unit of consumption is the prepared dish, not the isolated ingredient, any data-driven analysis that fails to account for the transformative effects of cooking is inherently limited. Yet, a critical knowledge gap exists due to the profound lack of systematic, quantitative databases for these complex dietary systems. This dataset offers a critical empirical foundation for analyzing sustainability trade-offs and developing culturally adapted dietary solutions. To address these limitations, we developed a comprehensive database of 3,302 dishes from six major Chinese cuisines, namely Shandong, Jiangsu, Zhejiang, Fujian, Hunan, and Anhui. This release is independent from our previously published dataset on Cantonese and Sichuan dishes, and we do not merge the two releases because harmonizing definitions and processing across datasets is outside the scope of this manuscript. The database was constructed by systematically integrating and standardizing recipes from traditional publications, online culinary platforms, and field surveys. Its core feature is an integrated, three-dimensional assessment for each dish. We quantified comprehensive nutritional profiles using national food composition tables. Multi-dimensional environmental footprints, specifically greenhouse gas (GHG) emissions, water withdrawal and consumption, and land use—were calculated using localized life-cycle assessment data and cooking-process energy models. Finally, standardized market costs were derived from extensive surveys of individual ingredient prices to reflect economic affordability. The guiding research question of this study is how a dish-level assessment balancing nutrition, environment, and economy can facilitate a sustainable planetary dietary solution, thereby offering an empirical foundation for culturally adapted dietary transitions. Methodology Recipe data acquisition and standardization In this study, a recipe dataset initially comprising 3,302 representative Chinese dishes is established. The data were collected from several sources, including traditional culinary publications38 and online platforms39. To ensure data quality and consistency, a rigorous, standardized processing pipeline was implemented. First, a data cleaning procedure was performed to exclude recipes with incomplete text or significant omissions in their ingredient lists or preparation steps. Subsequently, we utilized the DeepSeek-V3 model (via Alibaba Cloud DashScope API) to facilitate automated data extraction, with the generation temperature fixed at 0.2 to ensure deterministic outputs. This process utilized a prompt template (available in the FigShare code repository) designed to parse raw recipe text into structured JSON fields, including standardized ingredient weights and categorical classifications. To ensure data integrity, all automated results were subjected to manual curation to identify and rectify anomalous records, ensuring the final dataset reflects realistic culinary and nutritional standards. A critical step in our methodology was standardizing ingredient quantities. There are some vague quantifiers commonly found in Chinese recipes (e.g., “one tomato”, “a bunch of chives”), which were converted to standardized gram weights using an LLM-based automated script and subsequently verified through manual curation to ensure they aligned with typical culinary portions. Similarly, ambiguous terms for seasoning amounts, such as “an appropriate amount” or “a little”, were assigned specific standard weights. Following automated processing, a manual curation process was conducted to identify and remove anomalous records, such as recipes with ingredient quantities that far exceed typical household use. Furthermore, to ensure consistency among dishes, we standardized the names of all extracted ingredients. This process addressed the issue of regional variations in ingredient nomenclature across China, where a single food item may be known by multiple names. This terminological standardization process involved mapping the original Mandarin ingredient names to their technical counterparts in Japanese databases. This was essential for accurately matching the ingredients in our dataset with established external sources, namely the Chinese Nutrition Society’s Food Composition Database40 and the Life Cycle Assessment (LCA) database from the Advanced Industrial Science and Technology (AIST)41. This matching enabled the assignment of detailed nutritional and environmental impact data to each ingredient. Price estimation The market price for each ingredient was estimated through a multi-step process. To ensure temporal consistency, the base year for all market quotations was established as 2025. Initially, market quotations for a wide range of food ingredients in China were systematically scraped from e-commerce platforms42. Subsequently, the LLM-based model we established was utilized to match these prices with the standardized ingredients in our recipe dataset. For ingredients that could not be matched through this primary method, their average prices were collected from other major e-commerce platforms. Finally, all price data underwent a manual verification process to ensure accuracy and reasonableness. The total cost of a dish Ctotal was then calculated by summing the costs of all its constituent ingredients. The cost of each ingredient is the product of its unit price and its mass, as shown in the following equation: where pi is the unit price (e.g., CNY/kg) of ingredient i, and Mi is the required mass (kg) of that ingredient in the recipe. Nutritional components estimation The nutritional value of each dish was estimated by quantifying its constituent components. Standard nutritional content data for every 100 g of edible ingredient were primarily sourced from the Chinese Nutrition Society’s Food Composition Database40, supplemented with data from the Eshian Food Database43. In cases where data for specific ingredients were unavailable, values from nutritionally similar ingredients were used as a proxy. In total, 28 key nutritional indicators were estimated for each recipe, including: Energy (kcal); Protein (g); Lipids (including Fat, g; Saturated fatty acids (SFA), g; Monounsaturated fatty acids (MUFA), g; Polyunsaturated fatty acids (PUFA), g; Cholesterol, mg); Carbohydrates (including Total carbohydrates (CHO), g, and Dietary fiber, g); Vitamins (Carotene, μg; Vitamin A, μg; α-Tocopherol (α-TE), mg; Thiamin, mg; Riboflavin, mg; Niacin, mg; Vitamin C, mg); and Minerals (Ash, g; Calcium (Ca), mg; Phosphorus (P), mg; Potassium (K), mg; Sodium (Na), mg; Magnesium (Mg), mg; Iron (Fe), mg; Zinc (Zn), mg; Selenium (Se), μg; Copper (Cu), mg; Manganese (Mn), mg; and Iodine (I), μg). Environmental footprint accounting As an essential aspect of sustainable diet, the environmental footprints of food systems matter to planetary sustainability. To take the environmental load of each recipe into consideration, we quantified three dimensions of environmental footprint using a farm-to-table LCA framework (encompassing production, processing, transportation, and cooking). These included greenhouse gas (GHG) emissions (kg CO2eq per dish), water withdrawal and consumption (m3 per dish), and land use area (m2 per dish), as these indicators have already transgressed planetary boundaries due to anthropogenic activities44. Carbon footprint The carbon footprint evaluation measured the total greenhouse gas (GHG) emissions associated with a recipe’s ingredients throughout their life cycle, from agricultural production to the consumer’s table (farm-to-table). This included emissions from production, processing, transportation, and the final cooking process. We excluded transport from retail to home due to large household level variability. Carbon footprint embodied in food ingredients The embodied carbon footprint of each ingredient was calculated by matching our standardized ingredients with their corresponding emission factors from the AIST LCA database. The emission factors are primarily cradle to gate, covering resource extraction, transportation, material and energy supply, and production and processing up to the producer gate, while excluding downstream stages such as storage, packaging, retail, use, and end of life. This database provides global warming potential (GWP) values, converted to carbon dioxide equivalents (CO₂eq) based on the IPCC Sixth Assessment Report45. The total embodied carbon footprint of a dish Fcarbon_ingredients is the sum of the carbon footprints of all its ingredients, calculated as: where Mi is the weight of ingredient i (kg), and fcarbon_i is the emission intensity factor for that ingredient (kgCO₂eq/kg), derived from its corresponding GWP value. Carbon emissions from the cooking process Carbon emissions generated during cooking were calculated based on the type of appliance. Our model accounted for common cooking approaches in Chinese households, which were grouped into three types: open-flame cooking, electrically powered cooking (using appliances like ovens and rice cookers), and non-heated preparation. Open-flame cooking emissions. To model open-flame cooking, the natural gas stove was chosen as the representative appliance. This choice reflects its high prevalence in urban Chinese households, a result of extensive natural gas infrastructure and supply networks. The emission for natural gas was estimated based on the net calorific value (NCV) of 38.9 MJ/m³ and a carbon content (CC) of 15.32 tC/TJ, as reported by Shan et al. (2017)46. First, the NCV was converted to GJ/m³: Next, the CO₂ emission factor per terajoule (TJ) was calculated: Finally, the emission intensity per cubic meter of natural gas was determined: The operational parameters of the gas stove were based on the Chinese national standard GB 30720-2014, Minimum allowable values of energy efficiency and energy efficiency grades for domestic gas cooking appliances. We assumed a Grade 2 energy efficiency, corresponding to a thermal efficiency (η) of 59%. Gas consumption was calculated using the low heating value of natural gas (37.62 MJ/m³). To approximate realistic usage, we assumed an average power setting equivalent to 60% of the rated thermal load (261.9 kJ/min). The gas consumption rate Cgas_rate was calculated as follows: Using the derived emission factor, the average carbon emission rate for a gas stove operating at medium heat Egas_stove was calculated to be: Electric appliance emissions. Emissions from electric appliances were calculated using the 2022 national average carbon emission factor for the Chinese power grid, which is 0.5366 kgCO₂/kWh, as jointly released by the Ministry of Ecology and Environment and the National Bureau of Statistics47. For the cooking method using the appliance oven, given the absence of a mandatory energy efficiency standard for ovens in China and their fluctuating power usage, we adopted a representative model. Based on market research of common household ovens (800W–2500W), we selected an average power of 1650 W and assumed a 70% power utilization factor to account for operational fluctuations. The electricity consumption rate Celec_oven is: The corresponding carbon emission rate Eoven is: As for rice cooker, based on the standard GB 12021.6-2017, Energy efficiency limits and levels for rice cookers, and market analysis of common models (300W–1000W), we selected an average power of 650 W. The electricity consumption rate Celec_cooker is: The corresponding carbon emission rate Erice_cooker is: Following the standard GB 12021.2—2015, Energy consumption limits and energy efficiency levels for household electric refrigerators, we assume the energy consumption and emission from refrigerator usage. A reference power of 150 W representing the midpoint for a medium-sized household refrigerator is applied here. Hence, the total electricity consumption rate Celec_fridge is: The total carbon emission rate Efridge_total is: As refrigerator energy consumption is shared across all stored items and not solely dedicated to the ingredients of a single recipe, an allocation factor was applied. We assumed that refrigeration (as opposed to freezing) accounts for 40% of the energy use. Because the ingredients of a specific recipe represent only a small fraction of the items stored in a household refrigerator, 5% of this refrigeration energy was allocated to the storage of ingredients for a given recipe. The final cooking-related refrigeration carbon emission Efridge_cooking was therefore calculated as: Water footprint The water footprint was assessed within the LCA framework using detailed factors from the AIST LCA database. In this database, the assessment of water footprint considered three components of water use, namely withdrawal water (total water extracted from surface, ground, or rain sources), consumptive use (water lost via evaporation, transpiration, or incorporated into the product), and non-consumptive use (water that is withdrawn but returned to the environment). The total water footprint of a dish Fwater_total is the sum of the water footprints of all its ingredients, calculated as: where Mi is the mass (kg) of ingredient i, and fwater_i is the unit water footprint factor for that ingredient (m³/kg), which integrates withdrawal, consumptive, and non-consumptive uses. This aggregation is used to provide a single comparable indicator of overall dish-level water footprint that is consistent across all ingredients in the AIST LCA database. As the objective of this study is cross dish comparison of overall environmental burdens rather than stage specific water management, we report the combined indicator as the primary measure. Land footprint The land footprint quantifies the land area occupied for food production. The total land footprint of a dish Fland_total is calculated based on the mass of its ingredients and their corresponding land use factors: where Mi is the mass (kg) of ingredient i, and fland_i is the land use factor (m2/kg) for that ingredient. This factor represents the area of land required to produce 1 kg of the ingredient. It is calculated as: where Ai is the total area of new or continuously occupied land (m²) for producing ingredient i, Yi is the annual yield per unit area (kg/year), and Ti is the duration of land use for that production type (years). The land use factors were derived from the AIST LCA database, which references official statistics such as Japan’s Ministry of Agriculture, Forestry and Fisheries’ 2015 Crop Statistics and 2015 Livestock Production Cost Statistics48, with localization adjustments made for key ingredients in Chinese cuisine. Data Records All dataset files have been deposited in the Figshare repository and are openly available for download49. The dataset provides a multi-dimensional sustainability assessment of 3,302 dishes from six major Chinese cuisines: Shandong, Jiangsu, Zhejiang, Fujian, Hunan, and Anhui. It comprises a series of spreadsheets that catalogue recipes across a wide range of preparations, from northern-style meat dishes to southeastern coastal delicacies. All recipe data were finalized on July 31, 2025. Each dish record includes two integrated assessments. A comprehensive nutrition profile, including macronutrients and key micronutrients, was calculated based on the edible portion of each ingredient, with results provided in the “Nutritional_content.xlsx” file. Multi-dimensional environmental footprints were quantified, covering the land use, water consumption, and full life-cycle carbon emissions associated with each recipe, and are reported in the “Environmental_footprint.xlsx” file. The types and measurement units of all environmental and nutritional indicators are documented in the accompanying “Type and Unit of Environmental Footprints and Nutrients.xlsx” file. The overall structure of the released database and the corresponding contents of each file are summarized in Table 1. Table 1Database Structure and Description.CategoryFile NameMeaningGeneral InformationType and unit of environmental footprints and nutrientsDefines the types and measurement units of all environmental and nutritional indicators.NutritionNutritional_contentReports the six essential nutrients (Energy, Protein, Fats, Carbohydrates, Vitamins, and Minerals) of each dish across major Chinese cuisines.Environmental FootprintEnvironmental_footprintReports the environmental footprints (Carbon, Water, Land) and serving weights of each dish across major Chinese cuisines.PricePrice_contentReports the price of each dish across major Chinese cuisines. Database Structure and Description. Technical Validation To validate the reliability, accuracy, and applicability of the dataset, we implemented a multi-stage technical validation framework, combining internal consistency checks, cross-comparison with other research, and quantitative uncertainty assessment. This process was designed to confirm that the dataset is internally coherent and representative of the intended scope. We first compared our recipe-level carbon footprint estimates with values reported in multiple studies3,15,50–54. As shown in Fig. 1, the distribution of our results consistently overlapped with the interquartile ranges of these references, spanning both lower- and upper-bound scenarios. Notably, despite differences in study design, functional units, and geographic coverage, our median values and overall dispersion remained aligned with the broader body of evidence. This convergence suggests that our estimation approach captures the correct order of magnitude and reflects realistic variability in dish-level carbon footprints. Cross-study comparison of dish-level carbon footprints. We further evaluated the dataset by examining the distributions of carbon, water, and land footprints as well as price across multiple cuisines and dish types (Fig. 2). The ridgeline plots revealed clear and systematic patterns. According to the results in panel a, the environmental footprints and prices of six major Chinese cuisines reveal an internally coherent pattern. However, from the perspective of food ingredients, animal-based dishes (e.g., beef & mutton, poultry & eggs) clustered at the upper end of carbon and land footprints, consistent with their well-documented environmental intensity. Plant-based categories (e.g., soy products, vegetables & fungi) consistently exhibited lower impacts across all indicators. Besides, we found cross-indicator coherence that categories with high carbon footprints also tended to have higher land requirements. As for price, results suggested a modest higher price in animal-based dishes, in line with earlier reports that higher-cost animal proteins tend to carry greater environmental burdens. Internal validation for the distributions of footprints and price by cuisine and by dish type. Ridgeline density plots of carbon, water, and land footprints and price, stratified by six Chinese cuisines (panel a) and by dish types/ingredients (panel b). To validate the plausibility of the nutritional values, we conducted a per-capita consistency check. Across all 3,302 dishes, the average energy and protein contents per person were approximately 292 kcal and 25.6 g, respectively. The protein and energy contents showed a moderate positive correlation (r ≈ 0.38), consistent with the expectation that higher-energy dishes tend to contain more protein. Outlier screening (3σ criterion) identified only a small number of extreme cases (7 dishes for energy, 56 dishes for protein), representing a negligible fraction of the dataset. Taken together, these findings support the internal consistency and external plausibility of the nutritional estimates at the per-capita level. In addition, we examined per-capita nutritional distributions and average environmental footprints across the eight major Chinese cuisines (Fig. 3). To note, the results for Cantonese and Sichuan cuisines are from our previous dataset21. The butterfly histograms illustrate that both protein and energy values fall within plausible dietary ranges, while cuisine-level means highlight systematic differences consistent with culinary traditions. Distributions of nutritional intake and average environmental footprints across eight Chinese cuisines. All calculations were based on transparent ingredient-level mappings and standardized conversion procedures (yield factors, cooking adjustments, emission factors). Together with the publicly available scripts and source data, this ensures that the results are reproducible and can be readily adapted to alternative parameters or updated environmental databases. Usage Note The dataset provides a standardized foundation for diverse downstream applications in public health, environmental science, and policy modeling. Researchers can utilize the integrated nutritional and environmental profiles to develop multi-objective optimization models for sustainable diets or to conduct comparative assessments of regional culinary traditions. Furthermore, the data serves as an empirical resource for policymakers to design culturally adapted dietary guidelines that balance health outcomes with environmental boundaries and economic affordability. For practical use, users are advised to utilize the ‘Cuisine’ and ‘Dish Type’ categories within the spreadsheets to filter and aggregate records for specific regional or ingredient-based studies. Supplementary information read me Acknowledgements This research is funded by Japan Society for the Promotion of Science KAKENHI B (grant number JP24K03146 and 24K03152), KAKENHI C (grant number JP23K11542). Author contributions Y.L. conceived and designed the study and wrote the manuscript. Z.Z. and X.S. were responsible for organizing all datasets during the revision stage, supporting additional data processing, and preparing and submitting the revised version of the manuscript. N.N. and L.H. contributed to the preparation of the original manuscript, data and code formatting, and visualizations. F.C. and T.S. were primarily responsible for calculations and result verification. Y.H., L.X. and Y.Z. assisted in compiling raw datasets and metadata. S.T., Z.T. and K.L. contributed to preprocessing and quality control of the data. M.Z., X.X., Q.Z. and Y.Y. supported integration of datasets and coordinated technical communication. All authors reviewed and edited the manuscript. Y.L. and Y.Y. supervised the project and were responsible for checking, proofreading, and validating the results. Data availability The dataset has been deposited to FigShare49. Code availability All scripts used for dataset establishment and data validation are openly available in the FigShare repository49. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Yin Long, Xiaodan Shen, Ziyi Zhou . Supplementary information The online version contains supplementary material available at 10.1038/s41597-026-07432-8. References Associated Data Data Citations Supplementary Materials read me Data Availability Statement The dataset has been deposited to FigShare49. All scripts used for dataset establishment and data validation are openly available in the FigShare repository49.