Life cycle assessment of Tehran Municipal solid waste during the COVID-19 pandemic and environmental impacts prediction using machine learning

Abstract

Life cycle assessment and machine learning were combined to find the best option for Tehran's waste management for future pandemics. The ReCipe results showed the waste's destructive effects after COVID-19 were greater than before due to waste composition changes. Plastic waste has changed from 7.5 to 11%. Environmental burdens of scenarios were Sc-1 (increase composting to 50%) > Sc-3 > Sc-4 > Sc-b2 > Sc-5 > Sc-2 (increase recycling from 9 to 20%). The artificial neural network and gradient-boosted regression tree could predict environmental impacts with high R². Based on the results, the environmental burdens of solid waste after COVID-19 should be investigated.

Graphical abstract

1. Introduction

One of the most important public health services that needs immediate attention in a crisis is municipal solid waste management (MSWM) [1]. The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, which began in Wuhan, China, and quickly spread around the world [1], caused many changes in the economy, health, society, and culture. The spread of the COVID-19 pandemic caused various problems in most countries, and different organizations have provided guidelines to prevent the virus from further spreading [2]. The International Solid Waste Association (ISWA) has proposed to separate the recyclable materials and the long-term storage of household waste produced by COVID-19 patients [3]. The United States Environmental Protection Agency (USEPA) suggested moving the COVID-19 patients' waste to a sanitary landfill [4].

Waste quantity and composition have undergone various changes during the COVID-19 pandemic. Because of accessibility restrictions, quarantine, social distancing, the implementation of health protocols, and the widespread use of disposable devices and personal protective equipment (PPE), the quantity and composition of generated wastes have undergone various changes [5]. The study in Tunisia showed that online shopping during the COVID-19 outbreak increased the amount of recyclable waste and reduced organic waste. Applying periodical quarantines and buying food as needed has also caused a decrease in households' consumption and changes in their living habits [6]. Also, the implementation of restrictions and the closure of the institution's activities, as well as the remote working of employees, were the other reasons for fluctuations in the quantity of waste generated [7]. Another study in Morocco showed that the closure of communities during quarantine effects people's habits and the frequency of their purchases; therefore, municipal solid waste (MSW) generation has decreased from 11 to 4% in March 2020 compared to February 2019 [8]. Therefore, it is necessary to investigate the changes that have been made in the quantity and quality of solid waste production.

Due to the fluctuations in the quantity and composition of waste with the spread of COVID-19, life cycle assessment (LCA) is one of the practical tools that comprehensively evaluates the environmental impacts [9]. LCA appraises the environmental impacts related to the system, product, or process and has become an increasingly suggested method for decision-making in SWM systems [10]. Considering the variations in waste composition after the COVID-19 outbreak, the hypothesis of the study was that the destructive environmental impacts of Tehran's waste management have intensified. In this regard, the comparison of the MSW LCA of Bali, Indonesia, showed that the content of degraded organic carbon (DOC) decreased during the COVID-19 pandemic, from 0.121 to 0.058. Also, the total global warming potential has been considerably reduced from 1859.6 to 420.8 kg CO₂ kg-eq/day [11]. Yao [12] reported that the effect of traveling inhibition was observed in the impact categories of greenhouse gas emissions (−255 Mton CO₂-eq), energy use (−4.46 EJ), smog formation (−9.17 Mton O₃-eq), mineral and metal use (−16.1 Mton), and acidification (−226 kton SO₂-eq). According to the literature review and based on the contradictory results, it was necessary to investigate the effects of the changes in Tehran's solid waste caused by COVID-19 on human health and ecosystem quality.

The LCA raised awareness of the environmental consequences of any anthropogenic activity. It is important to develop a model to anticipate the environmental impact categories based on the input parameters of the MSWM [13]. Currently, various algorithms of machine learning (ML) have been applied to modeling energy consumption and environmental impacts [14]. ML algorithms have been used by LCA practitioners to develop surrogate LCA models that use product design parameters to estimate LCA results for new products. On the other hand, surrogate LCA appears to be used to represent ML algorithms that have been trained on LCA data [15]. An artificial neural network (ANN) as an extensive computational model was used for problem solving in diverse areas of environmental issues [16]. The previous studies used ML techniques for predicting the MSW generation, but the prediction of environmental impact categories in LCA was not considered as much. Moreover, despite the importance of COVID-19 on waste management, no research on the environmental impacts of baseline and proposed MSW scenarios has been conducted in Tehran, Iran. Based on the knowledge gaps, the present study aimed to compare the quantity and composition of the waste generated in 2019–2020 (before) and 2020–2021 (after the COVID-19 pandemic). In the second step, the number of personal protection wastes was calculated. In the third step, the environmental effects of the waste management scenarios were compared before and after the COVID-19 pandemic. In the fourth step, the environmental impacts of the proposed new scenarios were assessed to find the best strategy for the future epidemic. In the final step, the environmental impacts, avoided materials, and consumed energy were modeled using ML algorithms such as ANN and gradient-boosted regression tree (GBRT) to find the best prediction algorithms.

2. Material and methods

2.1. Study area

Tehran is the capital city of Iran, with a population of around 8.7 million; the city has the second-largest metropolitan area in the Middle East and is the 21st largest city in the world. According to statistics, over 7500 tons of waste per day were produced in Tehran before the COVID-19 pandemic. The generated waste was transported to the Aradkouh complex by 300 garbage trucks, traveling 3860 km round trip per day. The wastes were processed in several units, including recycling, composting, incinerating, and landfilling (more details in Supplement 1) [17]. The location of the study area is shown in the supporting file (Fig. S1).

2.2. Description of the alternative scenarios

In this study, the baseline scenarios before (Sc-b1) and after the COVID-19 outbreak (Sc-b2) were compared, and five alternative scenarios were developed based on the solid waste generated during the COVID-19 outbreak and evaluated using LCA. The schematic and system boundaries of the Tehran MSWM are provided in Fig. S2. Also, the proportion of each process in the investigated scenarios is shown in Table 1 . Under Sc-b1, mixed waste is collected from curbside containers and transported to the Aradkouh complex using semi-automatic garbage trucks. The mixed waste was processed through a material recovery facility (MRF), and about 11% of the recyclable materials such as paper and cardboard, plastic, glass, and ferrous and non-ferrous metals were separated. The residual waste (89%) after the recycling process was sent to a different process, including composting (17%), incineration with energy recovery (3%), and the landfill (69%). In Sc-b2, after collecting and transferring the mixed waste to the Aradkouh complex, 9% of recyclable materials were separated using the MRF station. The rest of the waste (91%) was sent to the same process used for waste management before the COVID-19 pandemic. About 18, 3, and 69% of the residual waste were processed through composting, incineration, and landfilling. The proportion of the proposed scenarios was presented in the supplementary file (Table 1).

Table 1.

Disposal scenarios for solid waste.

Scenario	Recycling (%)	Recyclable (%)	Composting (%)	Incineration (%)	Landfilling (%)
Sc-b1	11	Paper (4.85)	17	3	69
		Glass (1.81)
		Metal (0.56)
		Plastic (3.78)
Sc-b2	9	Paper (3.31)	18	3	70
		Glass (1.94)
		Metal (0.48)
		Plastic (3.27)
Sc-1	9	Paper (3.31)	50	3	38
		Glass (1.94)
		Metal (0.48)
		Plastic (3.27)
Sc-2	20	Paper (6.60)	18	3	59
		Glass (2.43)
		Metal (0.53)
		Plastic (10.44)
Sc-3	0	0	0	0	100
Sc-4	0	0	0	100	0
Sc-5	12	Paper (3.31)	35	10	43
		Glass (1.94)
		Metal (0.48)
		Plastic (6.27)

2.3. Life cycle assessment

The LCA methodology based on ISO 14040 and 14,044 comprises four significant stages: goal and scope definition, life cycle inventory (LCI), life cycle impact analysis (LCIA), and interpretation of the results. The description of each stage is presented below.

2.3.1. Goal and scope definition

The first aim of the present LCA study was to evaluate the environmental impacts of solid waste management in Tehran before and after the COVID-19 pandemic. The second aim was to compare the environmental effects of the proposed scenarios to select the best strategy for management of the waste in a future epidemic. The results are expected to be used by decision-makers, and public awareness should be raised about the necessity of waste source separation. The selected functional unit for comparing the alternative scenarios was 1 ton of generated MSW in Tehran. The system boundary (Fig. S2) includes waste collection from the residential areas and transportation to the Aradkouh complex, waste treatment alternatives including mechanical waste sorting, composting, incinerating, and landfilling, as well as all related consumed materials and energy.

2.3.2. Life cycle inventory

The life cycle inventory contains all inputs (resources, energy flow) and outputs (emissions, energy, wastes, and materials). The required data about waste collection and transportation and the quantity and composition of the generated waste were obtained through correspondence and interviews with experts from the waste management organization of the Tehran municipality. In addition, literature [18] was also used for collecting data for inventories. The Ecoinvent (version 3.8) database of SimaPro 9.3.0.3 was used for background data (Table S1). More details on LCI were presented in Supplement 2. The list of emissions during waste collection and transportation, sorting, composting, incineration, and landfilling into water, air, and soil was presented in Tables S2–S13.

2.3.3. Life cycle impact assessment (LCIA) and interpretation

The inventory analysis cannot reflect the potential environmental impacts of different used resources and released emissions; therefore, the LCIA step should be done. The aim of LCIA is to quantify and convert the environmental burdens related to the consumed resources and released emissions through different methodologies such as Eco-Indicator 99, CML 2001, ReCipe 2016, IMPACT 2002+, and etc. [19]. The ReCipe 2016 method was applied for environmental impact assessment due to the most updated methods in LCA, and it's an excellent implementation of cause-and-effect routes for the calculation of both midpoint and endpoint characterization factors [20]. This method gives us an overview of 18 environmental impact categories at the midpoint level (Supplement 3) and three at the endpoint level. Characterization factors (CFs) at the midpoint level reflect the cause-impact pathway and focus on single environmental problems, while CFs at the endpoint level reflect damage at three higher aggregation levels: human health, ecosystem quality, and resource scarcity [21]. Although normalization and weighting are not mandatory by the ISO standards on LCA but frequently used in practice to identify important impact categories, one can better understand the meaning of results by comparing them with more familiar references or solving tradeoffs between results [22]. Weighting facilitates decision-making, where the trade-offs between impact category results make it impossible to select the preferred solutions from alternatives [22]. In this study, weighting was used to select the best scenario for future epidemics.

2.4. Artificial intelligence (AI)

In contrast to natural intelligence, artificial intelligence allows a machine to behave like a human in a variety of contexts, most notably decision-making [23]. Machine learning, a subset of AI, is used to describe mathematical and statistical algorithms created to learn from existing information in order to improve future performance [24]. By using ML, the most useful attributes can be collected, and other attributes that might not have a substantial impact on the model's accuracy can be ignored. ML can be updated to offer suggestions or techniques for an optimization procedure. It can be used in a real-time decision-making process to find ways to enhance a system's performance over the course of its whole life cycle. As opposed to the entire LCA, this makes it particularly helpful in the design process [25]. SWM is one of the most complicated areas in environmental engineering due to the many technical, environmental, socio-economical, and legislative parameters involved in waste management programs; therefore, non-linear processes cannot model, optimize, and predict the impressive parameters. This study used ANN and a gradient-boosted regression tree (GBRT) to model the environmental impacts of Tehran's solid waste management.

2.4.1. ANN model

The environmental problems related to solid waste are typically hard to elucidate with a linear relationship [26]. As one of the reliable and fault-tolerant nonlinear models, ANN can find and predict the complex relationship between inputs and outputs [27]. On the other hand, ANN was extensively used in environmental areas due to its great self-learning capability and excellent accuracy in solving complex non-linear interactions without complicated mathematical rules [28]. In this study, Orange Software (version 3.32.0) as an open-source machine learning and data visualization tool was used to train, validate, and test the results of LCA with a back-propagation feed-forward method with a different activation function, number of neurons, and number of hidden layers [29]. The networks were built with eight input variables, such as material consumption (electricity, diesel fuel, machinery, and chemical products), and 29 outputs, including 19 environmental impact categories and ten avoided products such as fertilizer and electricity from different processes during solid waste management. The results of life cycle impacts arising from different waste management scenarios were imported into the supervised ANN algorithm. The datasets were randomly subdivided into three subsets, namely training (70%), validation (15%), and testing (15%). The activation function, solver, and maximal number of iterations were adjusted on ReLu, L-BFGS-B, and 200, respectively. The correlation coefficient (R²), root mean squared error (RMSE), and mean absolute error (MAPE) in Eqs. Eq. (1), Eq. (2), Eq. (3) were used to evaluate the performance of the ANN model.

$$R^2=1-\frac{{\sum }_{i=1}^n{\left({\hat{Y}}_i-Y_i\right)}^2}{{\sum }_{i=1}^n{\left(Y_i-\bar{Y}\right)}^2}$$ Eq. (1)

$$RMSE=\sqrt{\frac{\sum _{i=1}^n{\left(\widehat{Y_i}-Y_i\right)}^2}{n}}$$ Eq. (2)

$$MAPE=\frac{100}{n}\sum _{i=1}^n\left|\frac{Y_i-\widehat{Y_i}}{Y_i}\right|$$ Eq. (3)

Where n is the number of samples, $Y_i\text{,}$ ${\hat{Y}}_i$, and Ȳ _i is the observed value, the predicted value by model, and the mean value of the environmental impacts or avoided products.

2.4.2. GBRT model

The GBRT model combines the strengths of regression trees and boosting. The resulting additive regression model is made up of stage-wise fitted decision trees [30]. Several parameters in GBRT were altered to improve prediction performance, including the number of estimators, maximum tree depth, and learning rate. The maximum number of trees in the model is determined by estimators, whereas the tree depth specifies the degree of interaction between features. Another important parameter that impacts the weighting of each tree in the final model is the learning rate. A low learning rate will increase the number of trees employed and make results more consistent. To minimize the root mean squared error loss function, the final model linearly combines all decision trees, with a contribution to the overall model weighted by the learning rate [30]. The method used was extreme gradient boosting (xgboost); the number of trees, learning rate, and limit depth of individual trees were set at 100, 0.1, and 5.

3. Results and discussion

3.1. Comparison of Tehran's MSW quantity and composition before and after COVID-19 pandemic

Fig. S3 shows the amounts of waste generated in Tehran before and after the COVID-19 pandemic and the percentage of their changes over time. According to Fig. S3 (a), during the COVID-19 pandemic, waste generation in Tehran had different fluctuations. Waste production increased in the first six months of the pandemic compared to the same period before the pandemic, and in the second six months, waste generation decreased to the same level as before the pandemic. Based on Fig. S3 (b), the total waste production in March 2020 in Tehran has increased by 14% compared to March 2019. In April 2020, the waste production compared to April 2019 increased to 10%, and in the following months of May, June, July, and August, the number of waste products compared to the mentioned months in 2019 increased to 8, 2, 2, and 0.7%, respectively. With the continuation of the disease, waste generation decreased from 1.83 to 22% from September 2020 to March 2021. Despite numerous peaks of COVID-19 disease, the amount of waste generated in the second six months of 2020 has been lower than in 2019.

Initially, this increase in waste generation with the COVID-19 outbreak may be attributed to the increase in population, but the comparison of production waste statistics from 2015 to before the COVID-19 outbreak shows a significant decreasing trend in waste generation with population growth. Tehran waste generation declined from 7500 to 5000 tons per day over time, but with the COVID-19 outbreak in 2020, the generated waste increased to 5500 tons per day. The Tehran Waste Management Organization (TWMO) developed an executive and operational strategy titled “KAP” as a waste reduction plan to reduce waste production over time in response to population growth. The main aim of an innovative and sustainable KAP plan was waste reduction at the source and subsequently maintaining landfill capacity in the Aradkouh complex [17]. The changes in Tehran's MSW quantity in the first half of 2020 were increasing, while those in the second half were decreasing. Although the COVID-19 outbreak has reduced air and noise pollution and upgraded biodiversity, the impact of lockdown on waste management is alarming. Because of the stockpiling of PPE, the fear of the virus, and the uncommon production of waste in homes and health facilities, there seems to be a waste crisis [31]. The intensification of disposable products and panic purchasing have raised production and consumption, hence failing efforts towards waste reduction [32]. In addition, lockdown has influenced the recycling market, deteriorating the competitiveness of recycled plastics over virgin plastics due to low oil costs and low demand [32]. The change from eating out to home-delivered food in Thailand has raised the plastic waste from 1500 to 6300 tons per day [33], while Milan (Italy) has experienced a 27% reduction in waste generation [34]. Also, declining waste generation can be attributed to the closure of educational centers and some institutions, layoffs in various jobs, the unfavorable economic conditions of households, and the reactivation of waste pickers. Therefore, the fluctuation in Tehran's waste generation in 2020 can be related to the COVID-19 pandemic and its effect on people's lifestyles.

In addition to waste quantity, the waste composition may be changed. Tehran's waste composition before and after the COVID-19 pandemic was presented in Table S14. According to the information obtained from TWMO, the content of organic waste decreased from 69 to 61% in 2020, and the percentage of valuable and non-valuable waste increased from 15 to 14% to 18 and 19%, respectively. According to the information we have gotten from TWMO, one component that clearly changed in composition was plastic waste. The amount of plastic waste increased from 7.5 to 11% in 2020. Besides plastic, sanitary and textile wastes have also increased from 6.4 to 10% and 4.8–6%, respectively. The quantity and composition of Tehran wastes were compared with other studies, and the results are summarized in Table S15. The literature review showed the trend of changes in the waste composition during the COVID-19 outbreak was different and has not followed the same pattern in different countries [35]. Because lockdown, as the most important achievement of COVID-19, has an impact on people's lifestyles, eating habits, and waste generation [36], in addition to the mentioned items, the fluctuations also depended on the socio-economic condition, restrictions on transportation, type of shopping and purchase amount, proper storage of food, waste management plan, and lifestyle [37,38]. The organic wastes increased in Spain, New York, and Indonesia, while in Canada, Tunisia, and Tehran they decreased. A noteworthy point was the increase in plastic waste and recyclable materials, which is related to the widespread use of disposable and packaged items [[6], [7], [8]]. The pandemic is still far away, and therefore, it is necessary to monitor waste quantity and composition regularly in the communities.

3.2. Comparison of baseline scenarios at midpoint level

The environmental impacts of the baseline scenarios on Tehran's waste management that were performed before and after the COVID-19 pandemic are shown in Fig. S4. As you can see in Fig. S4, the destructive effects of Sc-b2 in all impact categories were greater than Sc-b1, and this may be related to the waste composition changes. Therefore, the hypothesis that the environmental effects of waste after the COVID-19 outbreak in Tehran are harmful is accepted. Based on Fig. S4, total emissions are the sum of indirect and direct emissions. Indirect emissions are related to the production of electricity and diesel at sources that are used in the waste management center. Compounds released into water, air, and soil during various waste processing operations are considered direct emissions. Both baseline scenarios have inhibition effects in the categories of global warming potential (GWP), terrestrial acidification potential (TAP), and fossil resource scarcity potential (FRSP), while the most destructive effects have been on the human non-carcinogenic toxicity potential (HnCTP), and land use potential (LUP). In other categories of impact, no significant effects are observed. Interpreting the most detrimental and inhibitive impact categories and their influential factors is provided below.

Global warming potential: people and ecosystem health are affected by GWP, which is measured in kg of CO₂ equivalents (21). Based on Fig. S4, the total emission values of GWP in both Sc-b1 and Sc-b2 scenarios were negative, which indicates that the used scenarios have prevented the emission of −2.82E+02 and −2.50E+02 kg CO₂-eq/t MSW, respectively. According to the inventory data, the most important factors for avoiding GWP were the recycling of nylon, non-ferrous metals, PET, and HDPE, with values of 44.5, 20.5, 9.2, and 9%, respectively. It was found that plastic recycling plays a major role in reducing global warming potential. Plastic recycling prevents its production, which reduces GWP. Jeswani et al. [39] reported that the GWP of recycling and chemically treating a ton of LDPE from mixed plastic waste was 124% lower than that of producing virgin polymer. According to the current study, plastic waste increased by 34% in 2020 compared to 2019, while GWP remained stable. Despite the increase in plastic in the COVID-19 outbreak, no significant changes have been made in their recycling rate; therefore, the majority of plastics end up in landfills. Due to the low degradability of non-recyclable plastics, their landfilling has a lower impact on global warming compared to incineration. Demetrious and Crossin [40] reported that greenhouse gas emissions of mixed plastics through gasification-pyrolysis and landfilling were 1.87 and 0.0151 kg CO₂-eq/kg waste, respectively. The higher GWP was related to the higher direct emissions and the electricity consumption of incineration. Also, Zaman [41] concluded that the GWP of the pyrolysis-gasification and landfill processes of the MSW were 1000.1 and 40.04 kg CO₂-eq/t MSW, respectively. These results could be associated with the GHG impact of the detrimental residue of the incineration process leading to productions of CO₂ and CO during the decomposition of final residue. Therefore, landfilling may be a better option for managing mixed plastic waste than the incineration process.

Terrestrial acidification potential: in addition to GWP, Sc-b1 and Sc-b2 have had an inhibitory effect on TAP. The total potential of TAP from Sc-b1 and Sc-b2 scenarios was −1.97E+02 and −1.95E+02 kg SO₂-eq/t MSW, respectively. The inhibitory cause of terrestrial ecotoxicity was plastic (64%) and paper (30%) recycling, and it was related to the prevention of NH₃ emissions because of recycling. Khoo [42] showed that plastic waste recycling has more environmental benefits for TAP compared with other waste management options. Also, Demetrious and Crossin [40] reported a lower TAP for landfilling mixed plastic waste compared with gasification-pyrolysis.

Abiotic resource depletion potential (ARDP): This impact category focuses on natural resources, including energy resources, crude oil, and iron ore, which are regarded as non-living. FRSP refers to the future average amount of extra oil to be extracted, and it is expressed in kg oil equivalent [21]. Fig. S4 shows that Sc-b1 and Sc-b2 have negative values in the FRSP category of impact and that the implemented scenarios have reduced fossil resource consumption. Plastic products are made from petroleum hydrocarbons, and the recycling of nylon, HDPE, and PET leads to a reduction in resource consumption. The inventory data show the most critical factor in reducing the impact of this category was related to the recycling of nylon, HDPE, and PET, with values of 41, 19, and 14%, respectively. Das et al. [43] reported that a circular plastics economy decreased GWP and FRSP impact categories. Fossil fuel consumption for the LDPE production from waste plastics (0.86 kg oil per kg LDPE) was significantly lower than the production of virgin plastic (1.98 kg oil per kg LDPE).

Human non-carcinogen toxicity potential and land use potential: both scenarios have the highest adverse effects on human health and ecosystem quality, respectively. Sc-b1 and Sc-b2 caused 4.78E+02 and 5.05E+02 kg 1,4-DCB-eq/t MSW, respectively, of HnCTP. The inventory data showed that the highest contribution of non-carcinogenicity was related to zinc (Zn) and arsenic (As), with 62 and 36% contribution, respectively. The highest concentration of Zn and As was observed in compost emissions, and, therefore, composting has the main role in the non-carcinogen impact category. Due to the lack of source separation in Tehran, these heavy metals enter the waste stream through electronic equipment, paints, pesticides, and batteries, and finally enter the environment through different processes, especially during the composting process. Implementing a proper source separation program can decrease the effect of this impact category. Yuan et al. [44] investigated the effects of three scenarios in the China MSWM on environmental impact categories. The results showed that the scenario, which comprises 40% incineration, 30% composting, and 30% landfilling, has the highest non-carcinogenicity impact. Land use includes the long-term use of land or changing the type of land use. Significant environmental outcomes of land use are the decreasing accessibility of habitats and the decreasing diversity of biological species. It is expressed in area-time (crop) equivalent (m²a crop-eq) [21]. According to Fig. S4, the values of LUP for Sc-b1 and Sc-b2 were 8.03E+01 and 6.25E+01 m²a crop-eq/t MSW, respectively. Considering that a large amount of MSW in Tehran is landfilled, this factor increased the required land.

In addition to comparing the baseline scenarios in the mentioned impact categories, investigating the weighted values in the damage categories is also helpful in understanding the results. The comparison of the weighted values of Sc-b1 and Sc-b2 in the endpoint levels (Table S16) showed that the destructive effects values of Sc-b2 were slightly higher than those of Sc-b1 in three damage categories (human health, ecosystem quality, and resource depletion). For more information, Fig. 1 (a and b) shows the effect of input parameters on the damage category. According to Fig. 1 (a), the most inhibiting agent of adverse effects on human health and ecosystem quality in Sc-b2 was related to the recycling of non-ferrous metals, nylon, PET, and paper and cardboard. In Sc-b2, the contribution of recycling nonferrous metals, nylon, PET, and paper and cardboard to mitigating human health damage was 32, 19, 18, and 9%, respectively. For mitigation of damage to ecosystem quality, the role of recycling nylon, paper and cardboard, non-ferrous metals, and PET was 35, 21, 15, and 9%, respectively. To avoid damage to resource depletion, the recycling of non-ferrous metal, nylon, HDPE, and PET with values of 36, 20, 12, and 12%, respectively, has the most effects. Recycling a valuable portion of solid waste reduces the need for raw materials, and finally, resource depletion is prevented.

Fig. 1.

Contribution of different input parameters to damage categories of Sc-b2, a) inhibitory role, b) destructive role.

According to Fig. 1 (b), the most destructive impacts on human health and ecosystem quality are related to direct emissions, i.e., emissions released from various processes on solid waste such as sorting, composting, incineration, landfilling, and transportation. On other hand, direct emissions were responsible for 74 and 68% of the negative effects on ecosystem quality and human health, respectively. The list of direct emissions through various processes is presented in Tables S2–S13. Various compounds that are emitted into the atmosphere (CO₂, CH₄, CO, SO₂, H₂S, C₆H₆, NMVOC, PM, Cr, Cd, Cu, dioxin, furan, and so on), the water reservoirs (COD, BOD, TOC, PAHs, As, Cd, Cr, Zn, Pb, nitrogen compounds, anions, and cations), and the soil (Al, Zn, Ni, Mo, Cr, Sb, and so on) are the main agents for adverse effects on human health and ecosystem quality. Consumed electricity in composting and incineration has the greatest impact on human health and ecosystem quality after direct emissions. In the category of damage related to resource depletion, the most destructive effects were related to electricity use in composting, incineration, and diesel consumption in recycling, with values of 46, 19, and 26%, respectively. The reason for the destructive nature of electricity consumption is due to the use of fossil fuels for electricity production. According to the ecoinvent database and data analysis, carbon dioxide (91%), methane (6%), sulfur oxides (1%), and other gases (2%) play the main roles in damaging human health, ecosystem quality, and resource depletion in Iranian power plants [19]. Any change in the fuel or energy used for electricity generation can alter the amount of these damages. About 84% of the electricity generated in Iranian power plants depends on natural gas, and only 14% of the electricity is supplied from hydropower plants [45]. Therefore, replacing renewable energy sources such as hydropower, wind, solar, and biomass with non-renewable energy sources for electricity generation can reduce indirect emissions and consequently reduce detrimental effects on human health, ecosystem quality, and resource depletion. Therefore, moving toward renewable energy will reduce indirect emissions and, thus, reduce the environmental impacts.

Considering that the scenarios implemented on Tehran's MSW before and after the COVID-19 outbreak did not change significantly, the results showed that the destructive impacts of the produced waste after the pandemic were greater, which is related to the changes that were made in the waste composition. The effect of waste composition on environmental impacts also showed that plastic waste recycling has the most inhibitory effect on all impact categories, while direct emissions through the processes carried out on waste along with electricity consumption have the most destructive effects on the environment. Bassline scenarios concluded that it is necessary that other countries evaluate the environmental impacts caused by quantitative and qualitative changes made to production waste.

3.3. Comparison of impact categories of proposed scenarios

Considering the destructive effects of the implemented scenario on Tehran's MSW, it is necessary to propose scenarios and examine their impacts and damage categories in order to find the optimum scenario with the least environmental impacts and implementation capability for future pandemics. Therefore, the proposed scenarios (Sc-1 to Sc-5) were compared with Sc-b2, and their impact categories are presented in Fig. 2 . According to Fig. 2 (a), the most destructive effects of the investigated scenarios were observed in the impact categories of TAP, terrestrial ecotoxicity potential (TEP), HnCTP, and LUP, while except for Sc-3, the inhibitory effects of some scenarios were seen in the GWP, TAP, TEP, LUP, and FRSP impact categories. These impact categories in the mentioned scenarios and their agents were discussed below.

Fig. 2.

Comparison of proposed scenarios a) based on impact categories; b) weighted environmental damages.

Comparison of TAP in the proposed scenarios: in the TAP impact category, Sc-4 (complete incineration) and Sc-1 (increasing composting from 17 to 50%) have the most destructive effects, with values of 1.23E+02 and 9.89E+01 kg SO₂-eq/t MSW, respectively (Fig. 2 (a)). The inventory data showed the main cause of TAP in Sc-4 was emissions of NOx and SOx during waste incineration, and in Sc-1, it was NH₃ and nitrogen oxides released through increasing the composting ratio. Therefore, it can be concluded that nitrogen and sulfur compounds have the most negative impact on acidification potential. Liu et al. [46] reported that incineration has numerous capabilities for acidification because of the release of high concentrations of SO₂ and NOx pollutants. A LCA of Turkey's MSW revealed that increasing the composting ratio to 77% indicates approximately the same trend for acidification due to the emission of NH₃ and nitrogen dioxide [47]. Yay [48] calculated the acidification for the incineration scenario to be 0.414 kg SO₂-eq per ton of MSW.

Comparison of ecotoxicity potential in the proposed scenarios: ecotoxicity has potential effects on freshwater, marine, and terrestrial ecosystems. Fig. 2 (a) showed that the highest ecotoxicity was observed in the terrestrial ecosystem. The highest TEP value was observed in Sc-4 (2.04E+03 kg 1,4-DCB-eq/t MSW), which was not comparable to the other scenarios. The inventory analysis showed the main factors for the terrestrial ecosystem were copper (Cu) and antimony (Sb), with a total contribution of 82%. This matter relates to the lack of source separation facilities in Tehran. Zaman [49] compared sanitary landfilling and incineration for the management of Sweden's solid wastes. The results showed the TAP value of the incineration process was higher than the landfill. The inventory analysis revealed that the primary pollutants emitted to the atmosphere by waste incineration were vanadium (V), Sb, Cu, and nickel (Ni).

Comparison of HnCTP in the proposed scenarios: the category of HnCTP was the only impact category where almost all investigated scenarios had harmful effects. Based on Fig. 2 (a), the order of intensity of the HnCTP impact category was: Sc-1> Sc-5> Sc-b2> Sc-2> Sc-4> Sc-3, with the main contributions of Zn (62%), and As (36%). The values of HnCTP for Sc-1, Sc-5, Sc-b2, Sc-2, Sc-4, and Sc-3 were 5.91E+02, 4.16E+02, 2.05E+02, 1.99E+02, 8.80E+01, and 1.14E+00 kg 1,4-DCB-eq/t MSW, respectively. The results showed that as the composting ratio increased, the value of HnCTP increased due to the release of toxic metals. Diaz et al. [49] reported that increasing the composting rate to 100% in MSWM in Canada increased the freshwater ecotoxicity potential (FEP), TEP, and HnCTP by more than 100%. This is mostly because of the metals that get into soils and water supplies when composting is done.

Comparison of LUP in the proposed scenarios: in this impact category of LUP, Sc-3 (complete landfilling) and Sc-b2 have the destructive effects, and the values of LUP were 4.25E+02 and 6.24E+01 m²a crop-eq/t MSW, respectively (Fig. 2 (a)). One disadvantage of the landfill process is the need for more land [50]. The literature review showed that landfilling had the largest contribution to LUP. For example, LCA on China MSW revealed that a scenario composed of enhanced recycling into five categories plus incineration, composting, and landfilling had the most significant contribution to LUP (1.06E+01 m²a crop-eq/t MSW) [44]. LCA in Karkow showed the least LUP for a scenario that composed of 32% recycling and 88% composting (−2.40E+01 m²a crop-eq/t MSW) [51].

Since the examined scenarios at the midpoint levels have different effects, the weighted values of these scenarios at the endpoint levels are shown in Fig. 2 (b) and Table S17 to make it easier to compare their results. Based on Fig. 2 (b), two crucial points are revealed: 1) In terms of overall damages, Sc-2 and Sc-1 have the most deterrent and destructive effects, respectively; 2) The proposed scenarios' most detrimental and inhibiting effects have been on the category of human health damage, with the most negligible effects on ecosystem quality and resource depletion. According to Fig. 2 (b), it was concluded that the best scenario with the least environmental impacts is that of Sc-2. Recycling reduced contamination and the amount of MSW compared to when waste was abandoned in the environment, composted, buried, or incinerated. The results of this study confirmed that toxicity factors for humans and the ecosystem are reduced by recycling materials, particularly plastic waste. If the scenarios are compared based on the weighted values in the three damage categories, the order of the proposed scenarios in terms of the environmental burdens is: Sc-1 > Sc-3 > Sc-4 > Sc-b2 > Sc-5 > Sc-2. Except for Sc-1, which had destructive effects, other scenarios had inhibitory effects on all three damage categories. Based on Table S17, the highest destructive impact on human health belonged to Sc-1 with a value of 6.03 pt/FU, and the least effect on human health is correlated with Sc-2 with a value of −13.69 pt/FU. Due to the lack of source separation, mixed collected wastes can be polluted with toxic constituents, and therefore, increasing the composting ratio can lead to the release of toxic chemicals into the surrounding environment and adverse effects on human health and ecosystem quality [52]. In the category of damage to ecosystem quality, Sc-1 with values of 0.04 Pt/FU showed the most destructive effect, but other scenarios had inhibitory effects on ecosystem quality. In the damage to resource depletion category, the most destructive effect was that of Sc-3 (0.001 Pt/FU), while other scenarios had an inhibitory effect. Richard et al. [53] reported that the scenario that consists of 14% recycling, 19% landfilling, and 67% composting in Arusha City, Tanzania, has the least economic cost and environmental burden in most impact categories. Their sensitivity analysis showed that refining diesel consumption, decreasing CH₄ emissions to the air, and increasing the recycling ratio of papers and plastics would reduce the environmental impacts.

More than 70% of the waste in Iran has been organic, and one of the proposed options for the management of production waste has been composting. The results of this study showed that the destructive effects got worse as the composting ratio got higher. One of the reasons is the release of heavy metals through the discharge leachate as well as the gas emissions released into the atmosphere during the composting process. Therefore, it is suggested to provide the necessary infrastructure for source separation to reduce the risks caused by these pollutants during waste treatment processes. Considering that in Tehran, source separation infrastructures are being implemented, which make it possible to achieve 20% waste recycling, it is suggested that after the implementation of these infrastructures, a re-evaluation of the environmental impact be done and the output results should be compared with the present study in order to select and implement a process with the least environmental effects.

3.4. Evaluation of the models' performance

Machine learning, a subset of artificial intelligence, is used to describe mathematical and statistical algorithms created to learn from existing information in order to improve future performance [24]. The reduction of data collection costs was one of the potential advantages of ML in LCA. By using ML, the most useful attributes can be collected, and other attributes that might not have a substantial impact on the model's accuracy can be ignored. ML can be updated to offer suggestions or techniques for an optimization procedure. It can be used in a real-time decision-making process to find ways to enhance a system's performance over the course of its whole life cycle. The process can be improved by using optimization techniques. As opposed to the entire LCA, this makes it particularly helpful in the design process [24]. The prediction results of environmental impacts through the investigated scenarios according to ANN and GBRT algorithms were provided in Table 2, Table 3 , while those for avoided materials and energy were presented in Tables S18–S19. Orange software was used to implement the ANN model and data mining. ANN was used extensively in environmental areas due to its great self-learning capability and excellent accuracy, which allowed it to discover the complex relationship between inputs and outputs without complicated mathematical rules [54]. Table 2 shows the statistical results of the ANN model in predicting the environmental impact categories for the MSW of Tehran. The results of R², RMSE, and MAEP for ANN were calculated. To predict the environmental impacts using ANN, the values of R² vary between 0.969 and 0.999 for the training stage, 0.947–0.999 for the validating stage, and 0.889–0.997 for the testing stage. As it is clear from the R² values, RMSE, and MAEP, the ANN model has been able to predict the best impact categories with a high R² and low RMSE and MAEP. The lowest R² in the stages of training and validating is related to the impact category of MEuP, and for the testing stage, it belongs to the LUP. To predict avoided materials and energy using ANN (Table S18), the R² values diverge in the 0.998 to 0.999 range for the training stage, 0.989 to 0.999 for the validating stage, and 0.926 to 0.997 for the testing stage. For the ANNs model, the feed-forward backward propagation method with the ReLu activation function and the L-BFGS-B solver were used with 200 iterations. An 8-9-11-29 structure was approved as the predictive ANN model for the impacts of MSW in Tehran, i.e., two hidden layers were adopted with eight neurons in the input layer, nine and eleven neurons in hidden layers, and twenty-two neurons in the output layer. Nabavi-Pelesaraei et al. [55] used the ANN model to predict the environmental impacts of Tehran's MSW recycling. The results revealed that the feed-forward back-propagation ANN models based on the Levenberg-Marquardt (LM) training algorithm with a 5-7-7-11 structure were the best topology to predict the environmental impacts of recycled materials. Based on their results, the R² values vary in the ranges of 0.926–0.978 for the training stage, 0.935 to 0.983 for the testing stage, and 0.934 to 0.971 for the validating stage, which proves the capability of the ANN model for predicting the environmental impacts of the recycling process.

Table 2.

The results of the ANN model (8-9-11-29) for the prediction of environmental impacts.

Independent variables	Statistical indices of items of ANN
	Train			Validation			Test
	R2	RMSE	MAEP	R2	RMSE	MAEP	R2	RMSE	MAEP
GWP	0.999	0.555	0.344	0.999	0.555	0.344	0.972	0.977	0.652
SODP	0.999	0.005	0.000	0.996	0.009	0.005	0.989	0.065	0.022
IRP	0.976	0.017	0.007	0.976	0.016	0.007	0.946	0.029	0.017
OFP, HH	0.999	0.004	0.003	0.999	0.004	0.003	0.974	0.538	0.138
OFP, TE	0.999	0.006	0.003	0.998	0.007	0.004	0.996	0.220	0.034
FPMFP	0.999	0.004	0.003	0.998	0.006	0.003	0.944	0.303	0.082
TAP	0.993	0.004	0.002	0.992	0.008	0.003	0.989	0.512	0.016
FEUP	0.989	0.013	0.002	0.984	0.003	0.002	0.978	0.299	0.074
MEUP	0.982	0.003	0.003	0.947	0.005	0.004	0.942	0.324	0.077
TEP	0.969	0.354	0.041	0.947	0.670	0.067	0.934	0.445	0.057
FEP	0.999	0.005	0.004	0.998	0.006	0.004	0.993	0.295	0.053
MEP	0.999	0.008	0.004	0.995	0.012	0.007	0.978	0.459	0.059
HCTP	0.999	0.004	0.003	0.999	0.022	0.007	0.959	0.537	0.176
HnCTP	0.998	0.516	0.342	0.994	0.783	0.459	0.925	0.947	0.105
LUP	0.997	0.020	0.013	0.992	0.020	0.015	0.889	0.943	0.123
MRSP	0.999	0.005	0.003	0.994	0.016	0.0007	0.974	0.020	0.013
FRSP	0.999	0.020	0.014	0.996	0.008	0.000	0.958	0.694	0.488
WCP	0.998	0.005	0.003	0.999	0.006	0.003	0.997	0.014	0.006
DEP	0.999	0.604	0.513	0.998	0.635	0.104	0.987	0.328	0.088

Table 3.

The results of GBRT for the prediction of environmental impacts.

Independent variables	Statistical indices of items of GBRT
	Train			Validation			Test
	R2	RMSE	MAEP	R2	RMSE	MAEP	R2	RMSE	MAEP
GWP	0.999	0.001	0.001	0.999	0.001	0.001	0.992	0.527	0.222
SODP	0.998	0.007	0.003	0.997	0.010	0.005	0.984	0.021	0.009
IRP	0.999	0.001	0.000	0.998	0.003	0.000	0.987	0.124	0.026
OFP, HH	0.999	0.002	0.000	0.999	0.003	0.000	0.985	0.126	0.030
OFP, TE	0.999	0.001	0.000	0.999	0.004	0.000	0.944	0.126	0.031
FPMFP	0.998	0.009	0.000	0.997	0.009	0.001	0.953	0.023	0.007
TAP	0.999	0.001	0.000	0.997	0.007	0.002	0.992	0.067	0.025
FEuP	0.999	0.004	0.000	0.998	0.008	0.001	0.958	0.321	0.021
MEuP	0.999	0.001	0.000	0.997	0.005	0.000	0.965	0.118	0.009
TEP	0.999	0.001	0.001	0.997	0.001	0.001	0.988	0.610	0.171
FEP	0.998	0.005	0.000	0.996	0.009	0.001	0.994	0.160	0.053
MEP	0.999	0.004	0.000	0.999	0.001	0.000	0.991	0.173	0.058
HCTP	0.997	0.001	0.000	0.995	0.001	0.000	0.967	0.530	0.155
HnCTP	0.999	0.001	0.001	0.999	0.001	0.000	0.953	0.663	0.176
LUP	0.999	0.005	0.000	0.999	0.002	0.000	0.995	0.021	0.009
MRSP	0.999	0.003	0.000	0.996	0.006	0.000	0.892	0.010	0.003
FRSP	0.999	0.001	0.001	0.997	0.003	0.000	0.984	0.130	0.042
WCP	0.998	0.001	0.000	0.999	0.001	0.000	0.995	0.270	0.064
DEP	0.999	0.003	0.000	0.999	0.005	0.000	0.986	0.695	0.043

The statistical parameters of GBRT at the three levels of prediction of environmental impacts are summarized in Table 3. According to Table 3, to predict the environmental impacts using GBRT, the values of R² vary in ranges of 0.998–0.999 for the training stage, 0.995 to 0.999 for the validating stage, and 0.892 to 0.995 for the testing stage. Based on Table S19 to predict the avoided materials and energy, the R² values vary between 0.998 and 0.999 for the training and validating stages and 0.943–0.995 in the testing stage. The lowest R² belongs to the paper and cardboard recycling (R² = 0.943). The results demonstrated that GBRT with 100 trees and a learning rate of 0.1 can predict environmental impacts and avoid materials with a high rate of R² and smaller values of RMSE and MAEP. These results show that an accurate prediction of the environmental impact of avoided materials and energy can be attained for the MSW of Tehran in the investigated scenarios and, thus, can be applied for future planning. The results of this study demonstrate that ML can be combined with conventional optimization techniques to improve their capacity to quickly discover environmental impacts [25].

3.5. The lessons learned regarding the application of ML with LCA

Based on the literature review [15,24,25] and the results of this study, the lessons learned regarding the application of ML with LCA were; i) this study used ML models for the prediction of LCA's environmental impacts. The results showed that some algorithms of machine learning can be applied to the optimization of input variables in LCA research. ii) one possible benefit of ML in LCA was that it could make it cheaper to gather data. iii) by using ML, the most useful attributes can be collected, and other attributes that might not have a big effect on the accuracy of the model can be ignored. iv) ML can be updated to offer suggestions or techniques for an optimization procedure. v) it can be used in a real-time decision-making process to find ways to enhance a system's performance over the course of its whole life cycle. vi) as opposed to the entire LCA, this makes it particularly helpful in the design process.

In conducting this study, some limitations were observed. One of the most important limitations of this study was the lack of data related to Iran except for electricity in the SimaPro library. To solve this problem, global libraries (GLO) were used, which are given in Table S1. Also, due to sensitivities regarding waste management in Aradkouh, the correspondence to obtain information was repeated several times. LCA is useful for establishing relative, not absolute, environmental sustainability; it can assist us in selecting which course of action is preferable, but it does not give us the ability to categorize actions as “sustainable” or “unsustainable.” [56]. Machine learning algorithms need large amounts of handmade and structured training data, and some of them are known as black-box models,” with the prediction accuracy related to specific contexts [25].

3.6. Main challenges and opportunities in post-COVID-19 for MSW management

During the COVID-19 outbreak, waste management around the world, including Iran, faced temporary problems. By learning from the experiences, threats can be turned into opportunities, and by eliminating weaknesses, a comprehensive plan can be proposed for future pandemics. Therefore, it is a unique opportunity to examine past shortcomings and make conscious efforts to develop technology and invest in waste services and infrastructure in Tehran. The first problem has been the lack of a comprehensive waste management plan in Tehran in emergency situations, which has led to the cessation of recycling activities at the beginning of the COVID-19 outbreak and concerns about the correct waste management option. Also, no specific guidelines were provided for managing the wastes that were produced by infected COVID-19 patients at home, while in the wake of the COVID-19 pandemic, various organizations, governments, and local authorities issued guidelines for waste management. The recent COVID-19 pandemic has underlined that MSW management services cannot continue as an unplanned emergency service. Official recognition of waste management services as an essential need can solve challenges such as a lack of attention to waste management infrastructure, an insufficient budget for waste management services, a poor maintenance culture, a lack of consideration for waste management in emergency situations, and the effects of economic crises on waste management services [57]. Preparing guidelines on how to manage infectious waste produced in similar conditions, which can be learned from the experiences of other countries, such as giving separately labeled plastics to households to collect infectious waste, and locating containers for collecting infectious waste next to municipal waste containers, is important. These measures can be brought into the waste management plan in emergency situations. Response measures and guidelines should be updated to manage waste generated in future pandemics.

Also, the other problems with waste management in Iran is the lack of a source separation system, which caused the entire waste to be landfilled without processing in the first two months of the COVID-19 pandemic. Due to the high transmissivity of SARS-CoV-2 through the infectious wastes and the different remaining active times of the virus on all kinds of materials and surfaces, these infectious wastes have created great environmental and health concerns for many communities. Therefore, implementing a source separation system and adhering to it are the best guarantees for the optimal recycling of produced waste [58]. The results of the LCA showed that by increasing recycling to 20% in Sc-2, harmful environmental and health effects were reduced. Also, the KAP plan, which started in 2018, has provided the background for the establishment of source separation facilities. Considering the application of artificial intelligence to environmental issues [59], its application to recycling is also being considered.

Considering the implementation of source segregation infrastructure in Tehran, cultivation should be done in this field. In addition to the activities carried out by the government, the active startups in Tehran's waste management have been able to help promote a culture of source separation. Considering the importance of separating and collecting infectious waste produced by patients at home, the municipality can launch startups for future pandemics to collect patients' waste to reduce the risk of transmitting pathogens through solid waste and also to carry out recycling activities with no disturbance. Due to the increase in plastic waste in COVID-19 and the lack of segregation, periodic training programs should be prepared for citizens and especially workers involved in waste in order to share the implementation plan on an international platform. The results of the study showed that the increase in recycling causes a reduction in emissions and also helps with the implementation of the circular economy in waste management and achieve the sustainable development goals (SDGs), especially SDG 12, which is related to the management of production and consumption [60]. Adopting circular economy models in the solid waste management sector not only makes it easier to transfer collected waste from disposal sites to recycling sites, but it also reduces waste generation and conserves more resources.

Iran is one of the developing countries, and the COVID-19 pandemic has shown that decision makers cannot ignore the needs of the waste sector by not financing research and assume to provide appropriate solutions in emergency situations. Despite the long history of waste management in Iran, over 70% of the waste produced in Iran is managed in unsanitary landfills [17]. Also, waste management has received less attention compared to other areas, and the studies conducted there are fewer. Considering the COVID-19 pandemic and the urgent need to obtain data and analyze the effects, it is necessary to provide more support to researchers so that the results can be a solution to future problems. The spread of COVID-19 and the urgent need for data collection and analysis caused more support for researchers in the field of waste management so that the results could solve future problems. Also, some of Iran's neighboring countries have a similar situation in waste management, and the results can be a guide for them to manage their waste in a suitable way in similar emergency conditions. Considering the acquisition of new experiences in waste management with the spread of COVID-19 and based on the suggestions, it is hoped that measures will be taken to create a comprehensive plan for future epidemics.

4. Conclusion

In this paper, the quantity and composition of Tehran's MSW were investigated before and after the COVID-19 pandemic, and a comparative LCIA study was conducted to propose the best scenario for waste management in future pandemics. The results showed that the COVID-19 outbreak has caused fluctuations in the production of MSW and compositions in Tehran. The organic waste decreased, while the plastic waste increased significantly. The evaluation of life cycle environmental impacts showed that despite the inhibitory effects of both baseline scenarios in all the impact categories, the inhibitory effects of Sc-b2 were less than those of Sc-b1. The LCIA results of the proposed scenarios indicate significant variations in the studied environmental impacts. In the investigated 18 impact categories, Sc-1 was the worst and most significant contributor to impact categories in terms of fine particulate matter formation potential (FPMFP), TAP, FEP, marine ecotoxicity potential (MEP), and HnCTP. In addition, Sc-3 with complete landfilling has the most harmful effects on the GWP, stratospheric ozone depletion potential (SODP), ionizing radiation potential (IRP), freshwater eutrophication potential (FEuP), marine eutrophication potential (MEuP), LUP, FRSP, and mineral resource scarcity potential (MRSP) impact categories. This study showed that Sc-2 significantly reduced environmental burdens, human health, and resource depletion when recycling increased from a baseline of 9–20%. The damage assessment results showed that Sc-2 is the least significant contributor to human health, while Sc-1, followed by Sc-3, were the most significant contributors in this damage category. Sc-3 and Sc-5 were the most and least significant contributors to ecosystem quality, respectively. The greatest resource depletion was observed by Sc-3. Therefore, it was concluded that the best scenario for managing Tehran solid waste in epidemic conditions is to increase the amount of recycling to 20%, which requires creating source separation infrastructure and equipping the MRF unit in Aradkouh. In addition to this scenario, the combination of composting, recycling, and incineration processes with values of 35, 12, and 10% is also recognized as the next option on Sc-2 regarding the lowest environmental impacts.

The feasibility of ANN and GBRT models was confirmed by the high rate of R² and the smaller values of RMSE and MAEP. The coefficient correlation of values for testing stages in the ANN and GBRT models were 0.926–0.997 and 0.943 to 0.995, respectively, which confirmed that both models had excellent performance in predicting all the output parameters. Finally, based on the LCIA results, it could be concluded that it is necessary to improve the current waste management plan for the MSW of Tehran. One of the most favorable measures is to increase the recycling rate to reduce environmental impacts and natural resource consumption. This action requires improving MSW infrastructure with more efficient recycling technologies. Given that over 20% of the dry materials in MSW in Tehran have recycling value, it is suggested that further research into various types of recycling systems such as curbside collection, drop-off centers, MRFs, and different percentages of recycling on environmental impacts be conducted. However, it was revealed that the COVID-19 pandemic caused disorganization in waste management at first, but it can provide an opportunity for conducting LCA to better address the environmental problems related to waste generation and to focus on the best management scenarios.

Author contributions

Sakine Shekoohiyan handled resources, conceptualization, methodology, validation, writing - original draft, writing - review and editing, formal analysis, project administration, and data collection. Mohsen Heidari contributed to conceptualization, methodology, original draft writing, review and editing, project administration, and advising. Mobina Hadadian and Homa Hosseinzadeh-Bandbafha participated in the investigation, formal analysis, writing of the original draft and its review.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Sakine Shekoohiyan reports financial support was provided by Tarbiat Modares University. Sakine Shekoohiyan reports a relationship with Tarbiat Modares University that includes: board membership and employment. The authors declare that the authors do not have any conflict of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article.

Data availability

Data will be made available on request.