Dish-level carbon and nutrition dataset for 4,403 Asian recipes

Abstract

Global dietary transition is becoming imperative as the booming food systems are degrading ecosystems, while malnutrition and diet-related chronic diseases are becoming increasingly prevalent. The choices individuals make about their diets affect their health through nutrition intake and have profound implications for planetary environmental sustainability. Asian cuisines, which are widely consumed within Asia and popular globally, play a pivotal role in this dietary shift. Hence, this study establishes a Pan-Asian culinary dataset comprising 4,403 representative dishes from 48 countries and regions, capturing the region-specific dietary diversity. For each dish, we quantified both environmental and nutritional outcomes. The environmental impact was assessed, encompassing the carbon footprint embodied in food ingredients and seasonings, along with cooking energy consumption, while nutrient composition was characterized through 21 macro- and micronutrients. This dataset provides essential insights at the dish level, thereby making generalized food recommendation guidelines more tangible and applicable to individuals.

Background & Summary

As global food demand surges to meet the needs of a growing population¹, projections indicate a staggering requirement of over 500 megatonnes (MT) of meat annually by 2050 for human consumption². This burgeoning demand sparks a critical debate on how to achieve concurrent advancements in human health and environmental sustainability^3,4,5. Previous analyses have underscored the adverse environmental impacts of red meat and other animal-sourced foods^6,7,8,9, steering recommendations towards plant-based diets for their health benefits, such as reducing blood lipids, and their lower environmental footprints^10,11,12,13. Conversely, research also points out that certain emission-intensive foods, such as seafood, offer unique nutritional benefits not readily available in terrestrial diets, including bioavailable micronutrients and essential fatty acids crucial for global food and nutrition security^14,15,16,17. Additionally, red meat is noted for its easily absorbable protein content, which is essential for human health^18,19. However, the complexity deepens as raw food materials undergo various cooking processes before final consumption, which can alter their nutritional value and environmental impacts^20,21. This complexity is often overlooked by the general public, leading to disproportionate consumption or avoidance of certain foods without considering individual nutritional needs that vary by region, sex, age, and physical activity level^22,23,24,25. Therefore, developing a nuanced, accessible, and scientifically backed dietary guide becomes paramount. This guidance should balance the environmental impacts of food production with the nutritional requirements of diverse populations, ensuring that dietary recommendations are both sustainable and tailored to meet the varied needs of individuals across the globe.

Asia’s population constitutes approximately 60% of the global total, with populous nations such as China and India collectively accounting for over 2.8 billion^26,27, so the continent’s culinary diversity mirrors its demographic significance. Asian cuisine, encompassing a wide array of food traditions like Chinese, Korean, Japanese, Indian, and Thai, is renowned worldwide for its remarkable variety and complexity^{28,29,30,31,32}. Among those, Chinese cuisine is particularly distinguished for its richness, characterized by the Eight Great Traditions, which reflect China’s diverse climates, historical developments, economic levels, and geographical features^33,34,35. This culinary richness is not static; it continually evolves through exchanges and fusions within modern Asian dining practices^36,37. As immigrants and various socioeconomic factors introduce new influences, these culinary traditions blend and innovate, sparking new trends that resonate globally^38,39,40. This dynamic interplay of cultures and cuisines not only enriches the global culinary landscape but also highlights the importance of incorporating regional food practices into discussions on global food security and nutritional strategies. The diversity and adaptability inherent in Asian cuisines underscore the potential for developing sustainable dietary models that can address both health and environmental objectives, reflecting the intricate relationship between local food traditions and global nutritional trends⁴¹.

In the pursuit of sustainable diets, an interdisciplinary strategy that combines agricultural science, nutrition, environmental science, and socio-economics is imperative^42,43,44. This collaborative effort is geared towards transforming agricultural practices to boost crop efficiency and promote low-carbon livestock farming while simultaneously encouraging consumers to shift from meat-heavy diets to more plant-based alternatives^45,46. The “farm to table” concept highlights the environmental impact disparities within the food supply chain—from production and processing to consumption—and underscores the importance of implementing sustainable practices to reduce these impacts^47,48. Transitioning towards plant-based foods, which are both nutrient-rich and have lower environmental footprints compared to animal-based products, can directly address the dual challenge of enhancing nutritional value and reducing the environmental costs of food production^49,50,51,52. This comprehensive approach underscores the necessity of integrating sustainable agricultural and consumption practices throughout the entire food supply chain to align with global environmental sustainability goals.

Given the profound impact of dietary choices on both personal health and planetary sustainability, our study aims to provide a detailed dataset comprising 4,403 representative recipes from 48 Asian countries and regions. Using a process-based life cycle assessment (LCA), we quantify each recipe’s carbon footprint, accounting for ingredients, seasonings, and cooking methods, as well as its nutrient composition, including 21 essential macro- and micronutrients. This granular, dish-specific analysis moves beyond broad dietary guidelines to deliver actionable insights for sustainable eating. By revealing the environmental and nutritional profiles of diverse Asian cuisines, our work provides a science-based foundation for shaping healthier and more sustainable food systems in the region and beyond.

Methodology

To analyze the carbon footprint and nutrition content of various Asian dishes, we followed a structured approach, as depicted in the flowchart (Fig. 1). This methodology encompasses several key stages: dish data collection, cleaning, estimation of nutrition content, estimation of carbon footprint, and establishment of recipe inventory.

Fig. 1

Flowchart of methodology for establishing Asian recipe inventory.

Dish data collection

To construct a comprehensive dataset of representative recipes in Asia, we primarily used the Cookpad dataset provided via the National Institute of Informatics (NII) Informatics Research Data Repository⁵³. For verification, we also retrieved region-specific recipe information from publicly accessible recipe websites such as ‘Recipe Encyclopedia of Ajinomoto’⁵⁴, ‘meishijie’⁵⁵, ‘allrecipes’⁵⁶, etc. To ensure the geographical representativeness of the recipe data, we referenced the population size of each country or region to determine the corresponding appropriate number of region-specific recipes. For instance, India and China, having the top two populations, were allocated the highest number of recipes. Conversely, for countries with smaller populations (such as Brunei), the recipe number was guaranteed to meet a minimum threshold of ten recipes. After the first round of search, we obtained over 15,000 Asian recipes.

To ensure the reliability and integrity of the constructed recipe dataset, we implemented further measures to filter and consolidate the raw data. By comparing the similarities in the main ingredients and cooking steps across recipes, we excluded duplicates or highly similar entries, ensuring that each dish included was representative of its country or region. For instance, during the selection process for Mongolian dishes, when both “roasted whole sheep” and “roasted lamb” were searched, only one was included in the dataset due to their overlap in primary ingredients and cooking instructions. This filtering process not only encompasses a wide variety of authentic regional dishes but also maintains a high standard of uniqueness and representativeness for each culinary tradition. For countries or regions with a large number of recipes, we also classified them based on geographical locations. Taking China as an example, we rigorously categorized 761 recipes into eight major cuisines following traditional Chinese culinary classifications: Cantonese (270 recipes), Fujian (78 recipes), Anhui (39 recipes), Hunan (40 recipes), Shandong (92 recipes), Sichuan (85 recipes), Jiangsu (35 recipes), Zhejiang (40 recipes), and other recipes (82 recipes) that did not fit neatly into the traditional classifications.

After the initial screening, we classified each recipe according to two complementary schemes. For Dish type 1, the classification is based on the dominant protein type of the main ingredient (Beef, Pork, Chicken, Seafood, Vegetables, or Other). Here, the ‘Other’ category primarily features lamb and other meats, along with a variety of staple-based dishes such as fried rice and noodles. In contrast, Dish type 2 is based on dish form (e.g., Meat Dishes, Seafood Dishes, Noodle-Based Dishes, Soups, Salads, Snacks/Desserts, Vegetarian Dishes, Beverages/Drinks, Side Dishes/Pickles, and Condiments/Seasonings/Sauces). This dual classification yielded a final dataset of 4,403 recipes drawn from 48 Asian countries and regions, representing the diversity of regional cuisines (Table 1).

Table 1 Recipe numbers collected from 48 Asian countries and regions.

In addition to the recipe ID, recipe name, and dish type, each recipe in our dataset provided comprehensive information, including ingredients (main ingredients, auxiliary ingredients, seasonings and condiments), directions, preparation time, cooking time, and servings. A detailed breakdown of the information captured for each recipe and its corresponding descriptive details is presented in Table 2.

Table 2 Detailed information and description of each recipe.

Dish data cleaning

Step 1. Ingredient name standardization

To standardize terminology and streamline subsequent calculations, ingredient names across the 4,403 recipes were harmonized. Descriptive terms specifying ingredient forms were removed, such as “chopped”, “melted”, “sliced”, “freshly”, “washed”, etc. Culinary terms for identical ingredients that vary due to regional linguistic differences or diverse etymological origins were consolidated. For instance, both “Zucchini” and “Courgette” refer to the same summer vegetable from the Cucurbita genus. “Aubergine” and “Eggplant” both denote the purple, typically elongated or oval-shaped vegetable. After removing duplicates, we ultimately obtained a list of 679 unique food ingredients.

Step 2. Ingredient unit standardization and weight conversion

The units and quantities of ingredients in each dish were converted into a standardized, calculable format. This conversion process encompassed two main aspects. The first part is the unification of ingredient units. For units expressed differently (in full name or abbreviation, singular or plural, etc.), we replaced them based on the meanings. For example, “tablespoon(s) of” was standardized to “tbsp”, “teaspoon(s) of” to “tsp”, and “liter(s)/litre(s)” to “L”. Small quantities such as “a little” and “a pinch of” and terms denoting non-edible uses like “for dressing”, “for topping”, “for garnishing”, and “optional”, were uniformly treated as 0 grams. For terms indicating unknown quantities, such as “appropriate amount”, “to taste”, “as needed “as required”, etc., we normalized them using the mean value of the corresponding similar ingredients. For cases without specific weights but only the number of ingredients, such as “1 potato”, “2 tomatoes”, “3 eggs”, etc., we utilized data from myfooddata (referenced from the United States Department of Agriculture, USDA) to convert these quantities into grams based on the large, medium, and small categories of the respective ingredients.

The second part of the conversion process involves the transformation of different ingredients from volumetric units to weight units (grams). Since the density of different ingredients varies, this conversion procedure is indispensable for accurately estimating carbon footprints and nutrient components of Asian dishes. The conversion coefficients we utilized were also sourced from the unit conversion tables from myfooddata. Here, we consider a range of common units, including tbsp, tsp, oz, lb, can, cup, slice, piece, fillet, sprig, leaf, sheet, etc. The complete coefficient table is provided in the GitHub repository⁵⁷ (file “Ingredient_weight_conversion.xlsx”).

For ingredient quantities given in volumetric units and specified by number rather than weight, the conversion to grams can be represented by Eq. (1) and Eq. (2), respectively:

$${W}_{i,j}={V}_{u,i,j}\times {\rho }_{i}$$

(1)

$${W}_{i,j}={N{um}}_{i,j}\times {{Average}}_{i}$$

(2)

where, ${W}_{i,j}$ is the weight of food ingredient $i$ in grams in recipe $j$, ${V}_{u,i,j}$ is the volume in the original unit $u$ for food ingredient $i$ (e.g., tbsp, tsp, cup), and ${\rho }_{i}$ is the density of the ingredient in grams per unit volume (e.g., grams per tbsp, grams per tsp, grams per cup). In Eq. (2), ${N{um}}_{i,j}$ is the number of food ingredient $i$ specified by count rather than volume or weight (e.g., “1 egg”, “2 tomatoes”), and ${{Average}}_{i}$ is the average weight of ingredient $i$ in grams obtained from myfooddata.

Step3. Revision of dish servings

The number of servings corresponding to each dish is key information for calculating the carbon footprint and nutrient components per serving. In most cases, we use the serving size exactly as declared in the source (Cookpad). Additionally, when a range is reported (e.g., 4–6 servings), we use the midpoint and include the range in the uncertainty notes. When neither information is available, we estimate servings by the total weight of the recipe. In the dataset we constructed, there were missing or unclear serving size data for a total of 462 recipes. For example, the dumpling dish provided the number of dumplings, the Baklava (a traditional Armenian dessert) provided the number of pieces, and the Katayef (a traditional Lebanese dessert) provided the number of slices. To minimize the error, we estimated the number of servings ${S}_{j}$ for recipe $j$ based on the total weight of ingredients in each dish, as expressed in Eq. (3).

$${S}_{j}=\frac{\mathop{\sum }\limits_{i=1}^{{n}_{j}}{W}_{i,j}}{{W}_{a}}$$

(3)

where ${W}_{a}$ is the assumed average weight per serving, ${n}_{j}$ is the number of ingredients in recipes $j$ without a known serving size. Notably, the typical grams-per-serving ${W}_{a}$ is empirically derived from recipes with declared servings, which can be found in the data repository.

Step4. Ingredient mapping

Each ingredient was matched with the most similar food item in the FAO-ASEAN Food Composition Database, which was systematically compiled from six ASEAN national food composition tables⁵⁸. The database clarifies food into different categories such as cereals, legumes, vegetables, fruits, meats, aquatic animals and products, etc. The food composition tables include the raw, cooked, and processed states of food, providing detailed information on the nutritional value of the edible portion per 100 grams. Despite the extensive coverage of this dataset, certain common ingredients prevalent in Asian recipes (e.g., steak, lamb shanks, cheese, sugar, bay leaf, etc.) were not covered. To address this gap, we supplemented the database with additional data from the FoodData Central (the USDA’s comprehensive source for food composition data). This ensured that common ingredients could be accurately matched for nutrient analyses. For detailed information, see the ‘Supplementary_nutritional_components.xlsx’ provided in the GitHub repository.

In this study, we conducted a comprehensive analysis of three major categories of nutrients (Table 3), where the main nutrients include energy (kcal), moisture (g), protein (g), fat (g), carbohydrates (g), dietary fiber (g), and ash (g); Minerals include calcium (mg), phosphorus (mg), iron (mg), sodium (mg), potassium (mg), copper (mg), and zinc (mg). In the realm of vitamins, we assessed retinol (mcg), β-carotene (mcg), vitamin A (mcg), vitamin B1 (mg), vitamin B2 (mg), niacin (mg), and vitamin C (mg).

Table 3 Nutrient types, abbreviations, and units of the considered nutrients.

Step 5. Dish categorization

We categorized the collected and refined 4,403 Asian recipes based on the main ingredients of each dish. The main purpose of this categorization was to systematically analyze the environmental impacts of different types of dishes, which may vary greatly due to their ingredient compositions and cooking methods. For each recipe, we separated the meat components, including beef, pork, chicken, and seafood, and classified those without meat as vegetable dishes. Following the initial classification, all dishes underwent manual verification to ensure accuracy. Ultimately, we divided the 4,403 Asian recipes into beef dishes, seafood dishes, pork dishes, chicken dishes, vegetable dishes, and other dishes (dish type 1 in Table 4). In addition, dish type 2 is a single-label culinary form/use category selected from a fixed taxonomy. We assign one label per recipe using an automated classifier (Alibaba Cloud DashScope) constrained to this taxonomy. When multiple cues are present, a deterministic priority rule favors canonical form/use labels so that the assignment remains single-label and reproducible. The taxonomy, priority rule, and the classification script are provided in the GitHub repository, and we performed a manual review to correct obvious misclassifications.

Table 4 Dish type count and the frequency.

Estimation of nutrition value

The estimation of nutrient components for each recipe is based on the weight of individual ingredient used and their respective nutrient components per unit weight. Each ingredient in the recipe is corresponded to the ASEAN Food Composition Database. Moreover, considering the parts that are typically removed and discarded during preprocessing and cooking, such as the skin and bones of meat, shells of seafood, roots of vegetables, and shells of eggs, we only calculated the nutrient components of the edible portion. The weight of the edible portion is determined using the percentage of edible portions listed in the Chinese Food Nutrition Component Inquiry Platform, combined with the ingredient quantities listed in the recipes. For each ingredient $i$ in a specific recipe $j$, the weight of the edible portion is calculated by:

$${W}_{i,j}^{{edible}}={W}_{i.j}\times {r}_{i}^{{edible}}$$

(4)

where ${W}_{i,j}^{{edible}}$ is the weight of the edible portion in ingredient $i$, ${W}_{i.j}$ is the total weight of ingredient $i$ listed in the recipe $j$, ${r}_{i}^{{edible}}$ is the percentage of the edible portion of ingredient $i$.

After obtaining the edible weight of each ingredient in one dish, the nutrient content ${N}_{i,j,m}$ for each nutrient $m$ in the edible portion of ingredient $i$ is given by:

$${N}_{i,j,m}=\frac{{W}_{i,j}^{{edible}}\times {N}_{i,m}^{{ASEAN}}}{100}$$

(5)

${N}_{i,m}^{{ASEAN}}$ is the nutrient content per 100 grams of the corresponding food item in the ASEAN Food Composition Database.

Finally, the per-serving nutrient content for the entire dish ${N}_{j,m}^{{serving}}$:

$${N}_{j,m}^{{serving}}=\frac{\mathop{\sum }\limits_{i=1}^{{n}_{j}}{N}_{i,j,m}}{{S}_{j}}$$

(6)

where ${n}_{j}$ is the total number of ingredients in the recipe $j$, ${S}_{j}$ is the number of servings of recipe $j$ obtained from Eq. (3).

Carbon footprint estimation for food ingredients and seasonings

To estimate each recipe’s carbon footprint, we account for the indirect emissions embodied in all its ingredients, including main and auxiliary ingredients, seasonings and condiments. The estimation is based on the quantity of each ingredient used (in grams) and its corresponding country-specific emission factor, reflecting cradle-to-gate life-cycle emissions, including production, processing, and transportation stages. To note, the emission factors ${f}_{i,k}$ for ingredient $i$ in country $k$'s recipes are derived National Institute of Advanced Industrial Science and Technology (AIST)⁵⁹, aligned with the IPCC 6th Assessment Report, which provides global warming potential (GWP) values converted to CO₂ equivalents⁶⁰.

To ensure region-specific accuracy, we use country-level data for ten Asian countries, namely China, Japan, South Korea, India, Indonesia, Vietnam, Thailand, Malaysia, the Philippines, and Singapore. In cases where country-specific data are unavailable, global average values are used as substitutes. This database ensures region-specific accuracy, reflecting their unique energy structure, production processes, and combustion efficiency⁶¹. By Eqs. (7) and (8), the carbon footprint for each dish and the per-serving carbon footprint are determined.

$${E}_{j,k}^{{\rm{indirect}}}=\mathop{\sum }\limits_{i=1}^{{n}_{j}}{W}_{i,j}\times {f}_{i,k}$$

(7)

$${E}_{j,k}^{{\rm{indirect}},{s}}=\frac{{E}_{j,k}^{{\rm{indirect}}}}{{S}_{j}}$$

(8)

In this equation, ${E}_{j,k}^{{\rm{indirect}}}$ denotes the indirect emissions of $j$-th recipe in country $k$, ${f}_{i,k}$ is the emission factor of ingredient $i$ per unit of weight in country $k$, ${E}_{j,k}^{{\rm{indirect}},{s}}$ represents the total emissions per serving for the $j$-th recipe in country $k$, and ${S}_{j}$ is the number of servings for the $j$-th recipe.

Estimation of Cooking CO₂ Emissions

To quantify the emissions attributable to cooking equipment, we follow a three-step procedure: (1) apportion total cooking time across different cooking methods; (2) compute energy demand and associated CO₂ emissions for each method; and (3) sum method-level emissions and normalize the total by servings. This approach builds on Long et al.²¹, adapting their thermal-efficiency framework to multiple cooking devices.

Step 1: Time allocation among cooking methods

For each recipe $j$, we recorded both the total cooking time ${T}_{j}$ (in minutes) and the specific cooking methods mentioned on the original recipe source websites. While some recipes involve only a single method (e.g., boiling), many traditional or complex dishes require multiple sequential or combined cooking techniques such as simmering followed by deep-frying, or steaming and then pan-frying. To handle this complexity and maintain consistency in estimating cooking-related energy use, we made a simplifying assumption: unless explicit duration breakdowns were provieded, the total cooking time was equally distributed among all identified cooking methods. This approach allows us to systematically apportion time across methods when detailed timing is not available.

Accordingly, the time allocated to each method $h\in {H}_{j}$ is calculated as Eq. (9):

$${t}_{j,h}=\frac{{T}_{j}}{|{H}_{j}|}\,\forall \,h\in {H}_{j}$$

(9)

where ${t}_{j,h}$ is the time allocated to method $h$ (min) and ${H}_{j}$ is the set of distinct cooking methods used in recipe $j$.

Step 2: Method-specific energy demand and CO₂ emissions

We distinguish two broad classes of cooking methods, namely electric‐device methods (${H}_{j}^{e}$), including microwave, oven, rice cooker, and thermal‐process methods (${H}_{j}^{t}$) covering boil, simmer, steam, deep-fry, stir-fry, and pan-fry.

Electric-device methods

The electrical energy consumed by method $h\in {H}_{j}^{e}$ is denoted as Eq. (10):

$${Q}_{j,h}^{{elec}}={P}_{h}\frac{{t}_{j,h}}{60}$$

(10)

and the associated CO₂ emission can be estimated as follows:

$${E}_{j,h,k}^{{\rm{direct}}}={Q}_{j,h}^{{elec}}\times {f}_{k}^{{elec}}\times {\xi }_{h,k}$$

(11)

where ${P}_{h}$ is the device’s rated power (kW), ${f}_{k}^{{elec}}$ is the country- or region-specific grid emission factor (gCO₂/kWh), and ${\xi }_{h,k}$ is an efficiency factor for different equipment (e.g. ${\xi }_{{micro},k}=1.4$ for microwave and ${\xi }_{h,k}=1$ otherwise).

To note, the appliance correction factor ${\xi }_{h,k}$ is used only for microwave ovens where cooking power is not the rated input power. In that case, we convert output-power-time to input-side electricity using ${\xi }_{{micro},k}$. Device power is taken from manufacturer specifications of representative Asiamarket models and listed in ‘Microwave oven spec.xlsx’ (in GitHub repository).

Thermal-process methods

Thermal‐process cooking operations, namely boiling, simmering, steaming, deep-frying, stir-frying, and pan-frying, are modeled as the sum of two distinct heat‐transfer components: the sensible-heat required to raise the temperature of the cooking medium (water or oil) from ambient to its operational temperature, and the latent-heat required to vaporize a fraction of that medium during the cooking period. In our framework, the sensible-heat demand for a given method $h$ is calculated as Eq. (12):

$${Q}_{j,h,k}^{{\rm{sensible}}}=\frac{{q}_{{\rm{med}}}{\times c}_{{\rm{med}}}\times \triangle {\rm{T}}}{1000{\eta }_{1,k}}$$

(12)

where ${q}_{{\rm{med}}}$ (g) is the mass of medium, such as water (for boil, simmer, steam) or oil (for deep-fry, stir-fry, pan-fry), ${c}_{{med}}$ is its specific heat capacity (kJ·kg^-1·K^-1, respectively), $\triangle {\rm{T}}$ is the temperature rise (from room temperature to 100 °C for aqueous methods or to oil-target temperature measured in K), and ${\eta }_{1,k}$ is the first‐stage thermal‐efficiency factor that accounts for heat losses to the cookware and ambient air.

Simultaneously, the latent-heat component captures the energy needed to vaporize water during cooking:

$${Q}_{j,h,k}^{{\rm{latent}}}=\frac{{{v}_{{\rm{eva}}}\times t}_{j,h}\times {H}_{w}\times S}{1000{\eta }_{2,k}}$$

(13)

where ${v}_{{eva}}$ denotes the evaporation rate (0.0621 g/s), $S=408.2$ and ${H}_{w}=2250{\rm{kJ}}/{\rm{kg}}$ are surface‐area and latent-heat parameters, and ${\eta }_{2,k}$ is the second‐stage efficiency factor reflecting additional losses during phase change.

By summing these two terms, we capture the total thermal-process energy demand ${Q}_{j,h,k}^{{\rm{thermal}}}$ for each method $h$ as Eq. (14):

$${Q}_{j,h,k}^{{\rm{thermal}}}={Q}_{j,h,k}^{{\rm{sensible}}}+{Q}_{j,h,k}^{{\rm{latent}}}$$

(14)

Based on the energy demand data calculated, we can convert it to the CO₂ emission by Eq. (15):

$${E}_{j,h,k}^{{\rm{direct}}}=\mu {\times Q}_{j,h,k}^{{thermal}}$$

(15)

where $\mu$ represents the fuel-combustion emission factor (also called the thermal energy emission intensity). In this study, $\mu$ is derived from the typical carbon intensity of natural gas combustion, as $\mu =\frac{13.6\times 44}{12}\approx 50\,{\rm{g}}{{\rm{CO}}}_{2}{\rm{per\; MJ}}$, and is consistent with IPCC default values for residential and commercial fuel use.

This two‐term heat‐transfer framework (Eqs. (12)–(15)) can be applied to all stovetop methods (e.g., simmer, steam, deep-fry, and stir-fry) by substituting the appropriate medium mass and thermophysical properties. For simmering and steaming, one replaces the generic ${q}_{{\rm{med}}}$ with the water mass and uses water’s specific heat and latent‐heat constants; for deep‐frying and stir‐frying, one uses the oil mass together with the oil’s specific heat and thermal transfer coefficient; and a fixed energy‐per‐batch value is assumed instead of computing both sensible and latent terms for pan‐frying.

Step 3: Aggregation and per-serving normalization

Having computed the CO₂ emission for each cooking method, the total cooking-equipment emission for each recipe is calculated by Eq. (16):

$${E}_{j,k}^{{\rm{direct}}}=\sum _{h\in {H}_{j}}{E}_{j,h,k}^{{\rm{direct}}}$$

(16)

To enable comparison on a per-portion basis, we then divide by the recipe’s number of servings as Eq. (17):

$${E}_{j,k}^{{\rm{direct}},s}=\frac{{E}_{j,k}^{{\rm{direct}}}}{{S}_{j}}$$

(17)

This framework yields a device-specific estimate of CO₂ emissions associated with cooking equipment, facilitating subsequent dietary footprint analyses.

Data Records

This dataset encompasses four inventories: the recipe count inventory (Recipe_counts.xlsx), the carbon footprint inventory (Asia_recipes_GHG.xlsx), the nutrition inventory (Asia_recipe_nutrition.xlsx), and the types of Chinese cuisines each recipe belongs to (Asia_recipe_Chinese_cuisine_type.xlsx), all of which are accessible on FigShare⁵⁷. It is designed to provide easily accessible information about the environmental impact, cost, and nutritional value of various recipes, thereby streamlining the research process for users. Utilizing the process-based LCA analysis, this dataset covers 26 fields, reflecting the embodied carbon footprint of ingredients, emissions from the cooking process, and the nutritional values of each recipe, detailing the main ingredients and serving sizes. To enhance user experience and facilitate precise data retrieval, Table 5 elucidates specific field names along with their associated data descriptors.

Table 5 Dataset structure and description.

Technical Validation

Distribution of Asian dish-level carbon footprint and nutritional value

We begin by examining sub-regional patterns across Asia (Fig. 2A). From a regional standpoint, Central Asia shows the highest dish-level carbon burden, averaging 734 gCO₂/serving, whereas West Asia (including Turkey, UAE, Syria, etc.) follows at 652 gCO₂, and East Asia (China, Japan, Korea) and Southeast Asia (Thailand, Malaysia, Indonesia, etc.) register more moderate footprints of 571 g and 562 gCO₂, respectively. South Asia (India, Bangladesh, Sri Lanka, etc.) lies just below the overall pan-Asian mean of 602 gCO₂, at 560 gCO₂/serving.

Fig. 2

Distribution of carbon footprint and nutritional value by country| Panel (A) Carbon footprint distribution (gCO₂eq/dish/serving). Panel (B) Nutritional content distribution. The unit of ENERC is kcal/serving. Units of WATER, PROCNT, FAT, CHOAVLDF, FIBTG, ASH are g/serving. CA, P, NA, K, FE, CU, ZN, THIA, RIBF, NIA, VITC units are mg/serving. And the units of RETOL, CARTB, and VITA_RAE are mcg/serving.

At the country level, the greatest average emissions occur in Kazakhstan, with 902 gCO₂ per serving per dish, more than twice the burden of East Timor, which has the lowest mean of 393 gCO₂. Other high-emitting nations include Azerbaijan (812 gCO₂), reflecting the prevalence of red-meat stews and energy-intensive cooking methods. In contrast, India and Sri Lanka average 457 g and 440 gCO₂/serving, respectively, consistent with their rice- and legume-centered diets. These disparities highlight the strong influence of regional ingredient choices and culinary techniques on the cradle-to-consumer carbon footprint across Asia.

The nutritional distribution of dishes varies across different countries (Fig. 2B). As for the energy content of dishes, East Asia shows the lowest mean at 402 kcal/serving, while West Asia leads at 611 kcal/serving, reflecting the meat-dense ingredients, common in countries like Qatar (the single highest country at 717 kcal). Central Asia and Southeast Asia fall in between, at 542 and 511 kcal, respectively, with South Asia averaging 389 kcal. In addition, protein, dietary fiber, and sodium are selected here as representative indicators to highlight regional differences in the nutritional distribution of Asian recipes. Protein trends similarly ascend in West Asia and Central Asia. Dishes from West Asia contain, on average, 33 g of protein per serving, topping the regional table, driven largely by Qatar (the highest country overall at 45 g). In contrast, South Asian dishes average just 18 g, with Sri Lanka having the lowest at 13 g. Dietary fiber is most abundant in West Asia (mean 5.1 g per serving), whereas East Asia and Southeast Asia average below 3 g. Fat content peaks in Central Asia at 27 g per serving, led by Kazakhstan’s 31 g, while East Asia dishes are leaner, averaging 11 g. Finally, sodium levels are highest in West Asia (2,430 mg per serving), while North Korean dishes show a high result at 3,529 mg, driven by the large amount of pickled foods, and lowest in South Asia (780 mg), exemplified by the Maldives at 317 mg. These disparities in macronutrient and mineral profiles across regions underscore the diversity of Asian culinary practices and their varying nutritional contents.

Then we analyze the nutritional content of different categories of dishes (Fig. 3). When recipes are grouped by Dish Type 1, beef dishes emerge as the highest‐emitting category with an average carbon footprint of 1,525.6 gCO₂/serving, more than twice that of vegetable dishes at 339.6 gCO₂/serving. Nutritionally, beef dishes also deliver the most energy (464 kcal), fat (22.6 g), phosphorus (314 mg), potassium (661 mg), iron (114 mg), and zinc (5.8 mg) per serving, in addition to elevated levels of protein (31.4 g), sodium (392 mg), and retinol (74.7 µg). In contrast, seafood dishes, with a footprint of 614.8 gCO₂, are the richest in calcium (432 mg) and maintain moderate levels of energy (320 kcal) and protein (23.2 g). Chicken dishes (486.3 gCO₂) excel in protein (26.5 g) and niacin (6.7 mg), while pork dishes (601.6 gCO₂) contain the highest levels of copper (0.25 mg), vitamin A (216 µg RAE), and thiamine (0.75 mg). The “other” category (699.0 gCO₂) encompasses staples like noodles and rice and shows a high carbohydrate content (49.2 g) and ash content (5.7 g).

Fig. 3

Distribution of carbon footprint and nutritional value by Dish Type 1 across all Asian countries| Panel (A) Carbon footprint (gCO₂eq/serving). Panel (B) Nutritional content distribution. The unit of ENERC is kcal/serving. Units of WATER, PROCNT, FAT, CHOAVLDF, FIBTG, ASH are g/serving. Units of CA, P, NA, K, FE, CU, ZN, THIA, RIBF, NIA, VITC are mg/serving. And the units of RETOL, CARTB, and VITA_RAE are mcg/serving.

By applying the more granular Dish Type 2 classification, we distinguish full meals from side dishes and beverages/drinks (Fig. 4). Meat dishes have the heaviest carbon burden (839 gCO₂) and the highest nutrient concentrations, with mean energy of 511 kcal, protein of 36.7 g, and fat of 32.1 g. Seafood dishes follow at 681 gCO₂, distinguished by peak calcium (210 mg) and sodium (1,200 mg). Among plant-based categories, salads and vegetarian dishes show low footprints (≈390 gCO₂) but high dietary fiber (8–9 g) and vitamin C (55–60 mg). Condiments/Seasonings/Sauces (226 gCO₂) and Beverages/Drinks (344 gCO₂) contribute minimally to calories (<310 kcal) and macronutrients, reflecting their auxiliary roles. Overall, both classification schemes reaffirm that high‐carbon dishes align with elevated energy and fat content, whereas low‐carbon categories provide superior fiber and micronutrient profiles at a substantially reduced climate cost.

Fig. 4

Distribution of carbon footprint and nutritional value by Dish Type 2 across all Asian countries| Panel (A) Violin-and-boxplot comparison of total cradle-to-consumer carbon footprint (gCO₂/serving). Panel (B) is the heatmap of nutritional content (energy and macro-/micronutrients) per serving (units are the same as in Fig. 3).

Sensitivity to time allocation across multiple cooking methods

In estimating emissions during the cooking process, when only one method and its cooking time are reported, we use them directly. Otherwise, we split the total cooking time equally among the methods. To quantify the potential error from this assumption, we conducted a perturbation sensitivity analysis. For each multi-method recipe, we took the equal-split result as the baseline emission and then randomly perturbed each method’s time share within ±30% of the equal share, while renormalizing to ensure the total cooking time remained unchanged. We generated 500 draws per recipe and compared perturbed emissions to the baseline using relative differences.

Across 1,579 evaluable multi-method recipes, the median absolute relative difference is 0.83%, the 90th percentile is 8.28%, and the median signed bias is approximately 0, indicating that errors are minor on average and roughly symmetric around the baseline. These results support using the equal-split rule as a reasonable fallback when per-method timing data are unavailable (see ‘cooktime_equal_split_perturbation_detail.csv’ in the GitHub repository).

Comparison with other results

To ensure the completeness and accuracy of the carbon footprint calculations for the recipes, we employed multiple databases in our validation. The validation approach included four datasets and involved data alignment. This was driven by the reason that some databases lacked information on the number of servings per dish. Hence, we analyzed the carbon footprint of entire dishes rather than per serving.

Results show that our dataset is in good alignment with previous datasets. Our dataset’s average carbon footprint per dish was 2,508.34 gCO₂eq, with a median of 1,712.21 gCO₂eq, a standard deviation of 2,809.99, and an interquartile range of 954.32-3,052.41 gCO₂eq. This broad distribution demonstrates the variability of Asian dishes, with values ranging from below 1 kg to over 10 kg CO₂eq per dish. In comparison, previous studies focused on Japanese recipes showed an average of 1,205.50 gCO₂eq, which is consistent with the relatively small portions and lighter cooking styles in Japan and aligns with our Japanese subset (1371 gCO₂eq on average)²¹. The other results for Japanese recipes contain only 45 recipes and show an average of 2,455.68 gCO₂eq¹². This might be partly explained by the exclusion of side dishes, condiments, seasonings, and sauces in this study. Besides, Yang et al.³³ reveal an average carbon footprint of 8,440.60 gCO₂eq and a median of 5,148.00 gCO₂eq in China’s dishes. Their sample comprises large, restaurant‐serving portions, whereas we model home‐style recipes with smaller, variable serving sizes. Yesildemir et al.⁶² quantified Turkish university lunch menus, reporting footprints similar to our results for Turkish dishes (2,257.5 versus 2,872.9 gCO₂eq), again reflecting dietary culture and portion size differences. Overall, our data distribution is similar to other datasets. This consistency indicates that our carbon footprint calculations are accurate and comparable with different databases.

Despite these advances, our study has several important limitations. First, we rely on emission factors from the IDEA database, which provides high-quality life‐cycle data for only ten Asian countries. As a result, recipes originating from regions outside this coverage must be approximated using global averages, potentially masking true national differences in production pathways and grid mixes. Second, our cooking-energy model applies uniform efficiency and grid-factor assumptions within each country, overlooking regional disparities in appliance use, fuel type, and kitchen infrastructure. In future work, we will extend our emission‐factor library with sub-national and commodity-specific data, refine cooking-energy parameters through direct measurement, and incorporate a broader diversity of culinary sources to further improve the geographic and methodological fidelity of our carbon footprint estimates.

Usage Note

This dataset covers 4,403 recipes from 48 countries and regions in Asia, each detailed in 26 indicators. It provides complete information on emissions and nutrition, specifying main ingredients and serving sizes.

For potential uses, individuals and families can utilize this dataset to select affordable, nutritious, and sustainable dishes, thereby improving their meal selections. Researchers in socio-economic science, environmental science, and nutrition can perform nutritional analysis and carbon footprint assessment. Data analysts and interdisciplinary researchers can explore the relationships between dietary habits, economics, health, and the environment. The catering industry can optimize menus and supply chains to offer dishes with lower environmental impact, higher nutritional value, and better cost-effectiveness, attracting consumers who prioritize environmental and health concerns. For policymakers in Asia, this dataset can help develop eco-friendly food policies and sustainable dietary guidelines that encourage the public to make healthier and more sustainable food choices. International organizations and decision-maker can also analyze dietary habits and their environmental impacts across different regions, providing a basis for global sustainable dietary strategies.

The advantages of this dataset are its extensive scope and multidimensional data. It includes a diverse range of typical Asian dishes, integrating detailed information on carbon emissions and nutrition, which supports interdisciplinary research and promotes sustainable practices among various populations. This dataset is publicly available under the terms of the Creative Commons License with Attribution (CC-BY 4.0).