Background & Summary

The fast development of the takeaway industry has been notable worldwide, with its share of sales of food services growing 11%-20% between 2009 and 20211. China boasts with the world’s largest user community for online food delivery, as nearly half (49.1%) of the country’s internet users have adopted takeaway into their daily routines as of 20232. This trend has driven the generation of takeaway packaging waste (TPW)3,4. It is estimated that China’s TPW generation in 2018 was nearly six times higher than in 20155,6. As food residues from Chinese food are not easily removed from the packaging, TPW is usually treated as general waste and ends up being incinerated4,7. The considerable volume of TPW puts new pressure on municipal waste management4,8. Due to a low recycling rate of plastic packaging4, TPW is risky to leak into the environment9. Plastic food containers, now among the top contributors to global river and coastal floating litter, can release microplastics and pose a risk to human health when ingested through the food chain10,11.

The Chinese government has launched the “Zero-waste City” program to promote effective and progressive local actions for municipal waste management12. This initiative necessitates targeted local policymaking to effectively control TPW. Studies have evaluated the generation of TPW and the associated environmental impacts at regional7,13 or city level6,14, highlighting the spatial variation in the TPW distribution. However, their considerations of takeaway packaging materials and usage probabilities are limited. They may underestimate the spatial heterogeneity caused by takeaway locations and urban structures (e.g., community characteristics and road networks). The area where takeaway restaurants are concentrated tends to consume more takeaway packaging and thus generate more waste after the packaging is discarded. Some previous studies15,16,17,18 enhanced spatial details by distributing aggregated data in a grid system, which informs the improvement of TPW spatial resolution. The grid zoning approach scales the data to a constant resolution, thereby eliminating the uncertainty caused by administrative boundaries, which can distort the underlying spatial patterns of TPW due to scale and area variance. Distributing aggregated data based on the spatial patterns of proxy data (e.g., population density and nighttime lights) can enhance spatial resolution, but its reliability requires further verification as it implicitly assumes a good spatial correlation between the proxy dataset and the aggregated data. Spatially distributing TPW based on the delivery scopes considers the process of TPW generation and could provide direct data to support source reduction and reuse of TPW.

To address these limitations, this study develops a scalable framework for estimating the TPW volume and mapping it with high resolution. This framework could also be applied to estimate packaging waste in express delivery services. A 1 km × 1 km grid dataset of TPW generation is constructed to support the effective management of TPW in Chinese cities. The volume of TPW at the street level is estimated using big data on takeaway consumption from the online food delivery industry in China19. Seven types of packaging and their usage possibility are considered. The grid dataset is available in GeoTIFF format for easy distribution and reuse.

This dataset could facilitate exploration in the following aspects. Firstly, it can be used to locate hotspots of TPW generation and potential risks of plastic waste leakage, helping local policymakers in tailoring effective municipal waste management. Secondly, the dataset can serve as input for life cycle assessments or other models, offering detailed information on waste generation and supporting detection of ecological risks related to packaging waste. Thirdly, the dataset can be used as panel data to analyze the socio-economic and natural drivers of TPW generation. This can help to identify patterns of TPW generation and distribution under different urban configurations, providing insights or experiences for TPW management in regions with similar urban structures.

Methods

Overview of the framework

The framework for constructing the grid dataset of TPW includes two steps (Fig. 1). First, the volume of TPW is quantified by multiplying the weight of packaging by the annual volume of takeaway consumption, using takeaway consumption data from the online food delivery industry. Next, the urban road network and the locations of takeaway restaurants are used to estimate the delivery scopes of the takeaway where TPW is spatially distributed, as takeaway packaging moves to the surrounding communities through delivery after they are packed in the restaurant. Gridded with geographic reference of administrative boundaries and water bodies, the high-resolution dataset of TPW is constructed with overlay analysis of the grids.

Fig. 1
figure 1

Framework for the construction of the gridded takeaway packaging waste dataset.

Full size image

Estimating takeaway consumption

The takeaway consumption data were derived from the primary online food delivery platform in China. The platform recorded the name and address of restaurants, takeaway consumption volume of restaurants of cities (over the past 30 days), and the type of cuisines. The takeaway consumption volume was scraped at the beginning of each month from March to August 2018 from https://waimai.meituan.com and were compiled in CSV format, and no individual information was involved. Due to the website redesigned, the URL no longer provides access to takeaway consumption data, and detailed information is now available on the mobile version of platform after authentication. The data was aggregated into nine categories: Chinese meals, local-flavor snacks, western food, Japanese & Korean cuisine, seafood & BBQ, noodles & dumplings, hotpot, and bakeries, desserts & drinks. Delivery of fruits, vegetables, medicine, tobacco, alcohol, flowers, and general merchandise were excluded through data cleansing, and the cleaned data covered 340,000 takeaway restaurants from 326 cities in China.

The average daily takeaway transaction volume between February 2018 and July 2018 reached 153 million and covered 88% of actual transaction of the online food delivery platform in 201820, indicating that the data was representative for TPW generation estimation. On the assumption that takeaway consumption volume was uniformly distributed over time, the six-month consumption volume was doubled to represent the annual volume6, and the industry-wide consumption volume was estimated by scaling up the consumption volume based on the market share21.

Calculating the takeaway packaging waste generation

Based on the takeaway consumption data of restaurants, the takeaway packaging waste at each takeaway restaurant was calculated using Eq. (1).

$${TPW}_{n}=\mathop{\sum }\limits_{i=1}^{I}\mathop{\sum }\limits_{j=1}^{J}({W}_{i}\cdot {N}_{ij}\cdot {P}_{ij}\cdot {Q}_{j})$$
(1)

where \({TP}{W}_{n}\) represents the sum of TPW generated at \({n}^{{th}}\) restaurant; \(i\) represents the type of takeaway packaging (i.e., food containers, chopsticks, toothpicks, napkins, drink cups, straws, packaging bags), with a total of \(I\) categories; j represents the aggregated takeaway consumption category; \({W}_{i}\) represents the weight of the packaging type \(i\); \({N}_{ij}\) represents the number of consumed packaging in type \(i\) for category j; \({P}_{ij}\) represents the use probability of the packaging type \(i\) for category j; and \({Q}_{j}\) represents the annual volume of takeaway consumption for category \(j\). In this study, the food containers are made of polypropylene, polystyrene, paper, or aluminum foil; chopsticks and the toothpicks are made of birchwood; napkins are made of bleached pulp; drink cups are made of polypropylene or polyethylene glycol terephthalate; straws are made of polypropylene; the packaging bags are made of low-density polyethylene. The weight of the packaging is sourced from our previous research6. The use possibility of packaging is sourced from Meituan Research Institute22.

Constructing the high-resolution gridded datasets of takeaway packaging waste in China

The takeaway food delivery scope of restaurants is configured based on the road network for spatially quantifying the TPW, as takeaway packaging is transferred from the restaurant to a nearby area through delivery service. To determine the food delivery scope of the restaurant, we performed a network analysis using ArcGIS Pro based on the locations of the restaurants and the road network extracted from an open map database OpenStreetMap (https://planet.osm.org)23 (Fig. 2). The intercity expressways are excluded from the road network, as most delivery vehicles (e.g., e-bikes) are not allowed on expressways24. Since 75% of takeaways were delivered within a 3 km radius24,25, the road network analysis was set with a cutoff distance of 3 km. A grid template was created by incorporating administrative boundaries and water bodies, extracted from the 1:250,000 national basic geographic information database publicly released by China’s National Geomatics Center (https://www.webmap.cn/commres.do?method=result25W)26, as the geographic reference.

Fig. 2
figure 2

Example of the derived delivery scope via road network analysis. (Expressways shown in light red lines were excluded from the road network analysis).

Full size image

The high-resolution TPW grid dataset was constructed by distributing TPW values onto a grid template based on the delivery scopes, as shown in Eq. 2.

$${{Grid}}_{m}=\mathop{\sum }\limits_{k=1}^{l}\frac{T{{PW}}_{k}}{{A}_{k}}$$
(2)

where, \({{Grid}}_{m}\) represents the value assigned to the \({m}^{{th}}\) grid of the template. There are \(l(l\ge 1)\) delivery scope(s) overlap at the \({m}^{{th}}\) grid. In this case, the grid value was assigned using aggregate sum resulting from overlay analysis of these overlapping areas (Fig. 3). For the \({k}^{{th}}\) restaurant, the distributed TPW value on the \({m}^{{th}}\) grid is the total \({TP}{W}_{k}\) divided by the geodesic area \({A}_{k}\) of its corresponding delivery scope.

Fig. 3
figure 3

Illustration of overlay analysis.

Full size image

Data Records

The dataset is made public on Figshare27, which is provided in two different formats for multidiscipline researchers. The first dataset is stored in GeoTIFF (.tif) format, providing a high-resolution (1 km × 1 km) grid of takeaway packaging waste for China in 2018 (TPWgrid-China). Data from Hong Kong, Macao, and Taiwan is currently unavailable due to limitations in data accessibility. Unit of grid values is tonnes (t), requiring multiplication by 0.01 for conversion to correct units. NoData value is defined as 65535. The dataset is georeferenced using WGS 1984 (WKID 4326), and projected using the Albers Equal Area Conic projection. The second dataset is organized in CSV format, providing detailed TPW accounting results for each city. The following fields are included: Province ID, Province Name, City ID, City Name, Total TPW, Mean TPW, Median TPW, and SD (standard deviation) of TPW. Both the Province ID and City ID follow the official coding standards from China’s Ministry of Civil Affairs28.

Technical Validation

Statistical analysis

Figure 4(a) shows the spatial distribution of TPW. In 2018, China generated 366.66 kt TPW. The top 10% of grids with the highest TPW generation accounted for 64% of the country’s TPW. Shanghai emerged top in the average TPW generation among cities at 27.48 t/km2, followed by Beijing, Chengdu, Fuzhou, and Nanjing, with respective average TPW of 15.56 t/km2, 14.41 t/km2, 13.50 t/km2, and 12.12 t/km2. Figure (b–e) shows enlarged views from four regions with high TPW production density: Beijing, Shanghai, Chengdu, and Guangzhou. Multiple aggregation centers of TPW emerged in these areas with road network characterized by gridiron patterns. In Beijing (Fig. 4b), for instance, eleven distinct hotspots of TPW generation marked the central urban area, including four aligned longitudinally, and three latitudinally. Corners of the urban area featured four hotspots on a smaller scale. The location of these hotspots is matched with the commercial or high-density residential areas. Nationally, China displayed marked disparities in TPW generation within the four major regions. In 2018, the eastern region, with 38.52% of the country’s population, generated 58.2% of the TPW, averaging 0.40 kg per capita. In contrast, the western region, which had 27.20% of the population and covered 71.56% of the territory, contributed only 16.48% of the TPW, with a per capita generation of 0.16 kg.

Fig. 4
figure 4

Spatial distribution of takeaway packaging waste in China (2018).

Full size image

The intensity of takeaway activities is linked to population size and economic vitality, which motivated us to examine the consistency of spatial trends between our dataset and these relevant datasets. Nighttime light data from remote sensing satellites is commonly used as proxy data for economic activity intensity29. Population density per unit area characterizes spatial patterns of crowd aggregation30. Five thousand locations were randomly selected for values extraction from TPW, nighttime-light data31, and population density data32. Spearman’s correlation coefficients show that TPW generation is moderately correlated with nighttime light illumination (0.38, p < 0.01), and there is a strong negative correlation between per capita TPW generation and population density (−0.71, p < 0.01). These results align with the fact that takeaway users are students and white-collar workers, who are typically located in areas with medium population densities, such as university campus, office buildings, and mid-end residential zones. Areas dominated by low- to middle-income groups tend to have higher population densities but lower consuming intention. Residents in these areas are less likely to order takeaway, leading to lower TPW generation in these regions.

Uncertainties and limitations

The origins of uncertainty in the dataset include two parts. Firstly, the estimation of the volume of takeaway consumption might introduce uncertainty. We accessed six months of takeaway records in 2018 and assumed a smooth flow of consumption volume throughout the year. This may disregard potential variations influenced by events such as public holidays or seasonal eating habits. Secondly, the estimation of the delivery scopes for TPW distributing may involve uncertainty. Due to the lack of detailed delivery data, we used the average delivery radius as the cutoff when performing road network analysis. This may underestimate the influence of interfering factors including weather conditions, traffic flow fluctuations, and the adjustment of delivery policy.

To reduce the uncertainty in the dataset, approaches can be considered from the following aspects: 1) Collect more comprehensive takeaway consumption data from broader sources covering a wider temporal span while refining the granularity of the data. This allows for a better capture of details of TPW associated with takeaway consumption, thereby enhancing the robustness of TPW dataset production. 2) Develop more accurate and detailed methods for TPW generation monitoring. A monitoring platform could be established for mapping and regulating the generation and disposal of takeaway packaging by incentivizing crowdsourced location marking of TPW generation, facilitating public participation, and ensuring trustworthy data sharing.

There exists a lack of comparable high-resolution TPW data. It motivated us to upscale the TPWgrid-China data to regional or national levels according to administrative boundaries, facilitating comparison with existing literature. In terms of spatial distribution patterns, TPWgrid-China is consistent with the findings of existing studies3,6. There is significant spatial heterogeneity in the TPW generation, which is mainly concentrated in the eastern provinces. Our estimated TPW volume is comparable to Huang et al.14 (846.68 kt) when considering a 126.5% increase in market size but significantly higher than Liu et al.7. We believe the differences stem from our study’s more comprehensive consideration of the packaging categories and consumption volume of the takeaway. We included beverage cups, straws, and napkins in addition to food containers, tableware, and packaging bags in the estimation. We took into account the probability and number of using different packaging categories and found that the average mass per PP plastic box was 4.5–11.5 g heavier than Liu et al.’s report.