Reduced health burden and economic benefits of cleaner fuel usage from household energy consumption across rural and urban China

Energy consumption in the residential sector is increasing rapidly in China. This study applies an integrated assessment model to investigate the adverse impacts of household energy consumption by various fuel types across rural and urban areas on age- and sex-specific premature deaths associated with PM2.5 pollution at provincial levels for 2015. We further estimate the economic and health co-benefits of a switch from solid fuels to electricity within households. We find that energy consumed by Chinese urban households is nearly 1.6 times than that of rural households. However, premature deaths due to household energy usage is 1.1 times higher in rural areas compared to urban areas due to direct use of coal for heating in rural households. The majority of household consumption-related premature deaths are predominately in the Southern area of China due to the population size and aging population. By replacing coal and biomass with electricity, this paper estimates economic benefits equal to 0.09% (95% CI: 0.08%–0.1%) GDP for rural areas and 0.006% (0.005%–0.007%) of GDP for urban areas of China. The results suggest that mitigation measures such as the promotion and subsidization of cleaner fuels, modern stove within rural households would yield these potential significant economic benefits.


Introduction
The residential sector is one of largest energy consumers in China (Fan et al 2013) and as such has a profound impact on the production activities, energy consumption and greenhouse gas emissions (Liu et al 2011). With increasing living standards and wealth across the Chinese population, residential energy consumption is forecast to continually grow in the short and medium term (Fan et al 2013). In China, solid fuels specifically coal and biomass (mainly wood and crop residues) are still important sources of energy for heating and cooking, largely in rural areas (Zhang and Smith 2007, Archer-Nicholls et al 2016, Yun et al 2020. Combustion of solid fuels by households cause both indoor air pollution (Zhang andSmith 2007, Clark et al 2013) and ambient air pollution at a local or regional scale . The emitted pollutants interact to produce a mixture of hundreds of different and hazardous chemicals known as secondary pollutants through physical processes and chemical reactions in the atmosphere .
Although household energy consumption is one of major anthropogenic contributors of atmospheric pollutants in China, studies on the impact of the household energy consumption on air quality remain limited (Du et al 2018, Zhao et al 2019, Yun et al 2020. The current evidence base has focused on elements of spatial and policy issues , Du et al 2018, 2019, Yun et al 2020 but has not fully accounted for the effect of household consumption on the health profile of different population groups at different spatial scales or across rural and urban areas. There is some evidence that while rural households have lower energy consumption compared to urban areas, they have higher rates of coal usage. And it is this element of fuel mix that may result in a higher burden for rural areas in terms of emissions and associated economic and public health outcomes. The difference in lifestyles, income, and population distribution between the rural and urban areas of China are pronounced (Hubacek et al 2009). Hence analyzing the impact of different activities and lifestyles across urban and rural populations, across different demographic groups may help to explain differential health outcomes (Pachauri et al 2004).
This study therefore undertakes an integrated assessment of the adverse impacts of household energy consumption by various fuel types across rural and urban areas on age-and sex-specific premature deaths as a key outcome from PM 2.5 pollution at the Chinese provincial levels for 2015. It does so through a modelling framework based on an existing integrated assessment model calibrated with data and submodels for population, spatial distribution of population, and air pollution loading. This paper then explores the health co-benefits and economic benefits of switching from coal and biomass to electricity since primary energy tend to be more and more transferred into electricity for usage in China.

Data and method
The definition of rural and urban households is based on China Bureau of Statistics definition, whereby urban households are those who have been living within the governance of a village (xiang) or town authority over 1 year and urban households are those who has been living in those areas where local governments of the county level or higher are located (www.stats.gov.cn/tjsj/ndsj/2015/html/zb06.htm). Land use in China is predominately rural: 91% of Chinese territory defined as rural (figure S1 (available online at stacks.iop.org/ERL/17/014039/mmedia)), excluding Taiwan, Hongkong and Macao). The analysis here focuses on households-these are, following standard definitions, individuals or groups of resident individuals who share the same living accommodation, and consume goods and services collectively (Ding et al 2017). 56.5% of the total population of China lived in 9.5% of the territory of China, with 775 million urban households and 596 million rural households (excluding Taiwan, Hongkong and Macao) (NBSC 2016). Sub-national data is provided for 30 provinces, cities and autonomous regions (Tibet is excluded due to a lack of energy inventory data). Fifteen fuel types are included in this study. These include electricity, gasoline, raw coal, heat, natural gas, liquefied petroleum gas, diesel oil, briquettes, other energy, other washed coal, coke oven gas, other gas, kerosene, fuel oil and coke. Heat as a form of energy usage, is additional energy produced by primary energy, such as coal combustion. Heat as a form of energy usage is predominately the result of coal combustion in China and is recorded in the Chinese energy balance table and as part of the energy inventory collected by China Emission Accounts and Datasets (CEADs, www.ceads.net). All data is from 2015 unless otherwise specified.

Household energy consumption
The calculation of household energy consumption is shown below: (1) E f,r,i, t refers to sum of energy consumption from each fuel type f, from household in urban or rural area r, at provincial level i, in year t. The energy unit is petajoule, PJ. AD f,r,i,t refers to the activity data of the combustion volume of each fuel type f. NCV f represents the net caloric value, that is heat value produced per physical unit of fossil fuel combustion. Here, data for the AD f,r,i,t is taken from the energy inventory data from China Emission Accounts and Datasets (CEADs, www.ceads.net), which is collected from China Energy Statistical Yearbook. Data of NCV f is taken from Shan et al (2018).

Pollutant emissions and PM 2.5 concentration from household consumption activities
Estimates from the household energy consumption model (equation (1)) are treated as the input to the Greenhouse Gas and Air pollution Interactions and Synergies (GAINS)-ASIA model. The GAINS model includes all key emission sources, with approx. 2000 end of control options (Amann et al 2011). The model is applied for estimating air-pollutants and PM 2.5 concentration based on E f,r,i,t (PJ). Emissions are calculated through a combination of three data categories: activity data, uncontrolled emission factors, the removal efficiency of emission control measures. Equation (2) represents the emissions estimates: where E p,i represents emissions of pollutants p, in province i; A a,i is the activity level of type a (e.g. coal consumption in residential sector) in province i; ef a,m,p,i is the emission factor of pollutant p for activity a in province i after application of control measure m; X a,m,p,i is the share of total activity of type a in province i to which a control measure m for pollutant p is applied. The resulting emissions are inputted into an atmospheric dispersion model, the EMEP Chemistry Transport Model (http://webdab.emep.int/) to compute annual mean ambient PM 2.5 concentration. GAINS employs reduced-form source-receptor relationships that have been derived from the EMEP atmospheric chemistry-transport model with a spatial resolution of 0.1 • × 0.1 • (Amann et al 2020). The PM 2.5 concentration are defined via: where C(PM 2.5 ) i is the PM 2.5 concentration in grid cell i. Em (PPM) represents the total primary PM 2.5 . The constantsπ, σ, α, β, γ are the source-receptor matrices for the corresponding pollutants contribution to the PM 2.5 concentration and the constants µ i are grid cell specific. More details of GAINS model can be found in Amann et al (2011).

PM 2.5 related health impact assessment
We consider the long-term exposure to PM 2.5 concentration on mortality as measured by premature deaths using the Global Exposure Mortality Model (GEMM) (Burnett et al 2018). Developed by Burnett et al (2018), the GEMM assess excess mortality attributable to ambient air pollution on a global scale.
Here we focus on the five leading causes of the PM 2.5 -related premature mortality: ischemic heart disease (IHD), stroke, chronic obstructive pulmonary disease (COPD), lung cancer (LC), and lower respiratory infections (LRI). The premature deaths under scenarios are measured by sex (female and male) and age group (25-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64, 65-69, 70-74, 75-79, 80+). For the purpose of this paper, we refer to the 'elderly' are individuals aged 65 years or more, in line with the World Health Organization definition (Orimo et al 2006, WHO 2010. The number of health outcomes are estimated by multiplying the relative risk (RR) with the population by sex and age and reported cause-specific incidence rate by sex and age, along with the corresponding uncertainties (95% confidence interval (CI)) and normally the percentage of exposed population is assumed to be 1 (Zhang et al 2017b): where M e,a,g,s,t is mortality in grid cell, k for each health endpoint, e by age, a and sex, g in year t due to PM 2.5 . RR k,e,a is the RR of a given PM 2.5 concentration in grid cell k at health endpoint e for age specific a, which is obtained from the GEMM (Burnett et al 2018). I e,a,g,t is the mortality rate for health endpoint, e by age, a and sex, g in year t obtained from the GBD Results Tool (http://ghdx.healthdata.org/gbd-resultstool) (figure S2). P k,a,g,t is the exposed population for age, a and sex, g in grid cell k, in year t. The spatial distribution of the population as a 0.1 • × 0.1 • grid in China in 2015 is taken from Xu (2017). Population data (NBSC 2011) and death rates (Chen et al 2020) by age group (5 year-old segments) and sex for 2010 are obtained at the provincial level. Following similar research, PM 2.5 intake is assumed to be equally harmful irrespective of the PM 2.5 composition and source and fuel of origin (Liu et al 2017).
The GEMM is expressed as: where RR e,a (C i ) is the RR of a given PM 2.5 concentration in grid cell i for age, a and health endpoint, e. C 0 is the threshold PM 2.5 concentration below which there is no additional risk, in terms of counterfactual PM 2.5 concentration (2.4 µg m −3 in this study).
And θ e,a , α e,a , µ e,a and V e,a are parameters describing the overall shape of the concentration-response curve, provided by Burnett et al (2018).

Attribution of premature death to household consumption activities
As per the GBD (2016), given the nonlinear relationship of the GEMM functions, the direct proportional approach was used to estimate premature deaths attributed to the emissions related to a region's production and consumption. The direct proportional approach assumes that the health impact of one pollution source is directly proportional to its contribution to the ambient PM 2.5 concentration. This proportional approach has also been applied to estimate the pollution health impacts related to household cooking (Chafe et al 2014), coal consumption (GBD 2016), road transportation (Anenberg et al 2017), and international trade (Zhang et al 2017a). For a given region, premature deaths due to different activity-based emissions (e.g. household consumption activities) can be calculated by multiplying the contributions of each to baseline ambient PM 2.5 concentrations by the total PM 2.5 -related mortalities for each grid cell.
M r,e,a,g,t is the premature death of household consumption of rural or urban regions r for health endpoint e by age, a and sex, g in year t. M k,e,a,g,t,base is the premature death of baseline scenario for health endpoint e by age, a and sex, g in year t in grid cell k. The baseline scenario uses modeled PM 2.5 concentration is taken from the GAINS model which has been validated against PM 2.5 monitoring data. Please see www.gains.iiasa.acat for more information. C k,t,base is the modelled PM 2.5 concentration for grid cell k in the baseline scenario in year t. C k,t,r is the modeled PM 2.5 concentration of grid cell k in alternative scenarios where the emissions related to household consumption of rural or urban regions r in year t.

Avoided premature deaths of replacing coal and biomass fuels into electricity
For the scenario analysis, we assume that households replace all coal and biomass fuels including cleaned coal, other washed coal, briquettes, coke, coke oven gas, other gas, other coking products and other energy (large part is biomass) with electricity but the total amount of household energy consumption remained the same. The GAINS model was used to calculate the number of avoided premature deaths across urban and rural area if current levels of coal and biomass consumption shifted to electricity consumption: where A r,e,a,g,t is avoided premature death by rural or urban populations, r for health endpoint e, by age, a and sex, g in year t. R k,e,a,g,t,r is the premature death attributable to household energy consumption without coal and biomass consumption but more electricity consumption of rural or urban regions r for health endpoint e by age a and sex g in year t in grid cell k.  figure S3). The calculation of economic benefit is shown:

Economic benefit of replacing coal and biomass fuels into electricity
where B is the economic benefit value of household consumption of rural or urban regions r in grid cell k for each health endpoint, e by age a and sex g in year t. VSL a,g,i,t is the value of a statistical life for age a and sex g in province i in year t.

Household energy consumption
In 2015, the total household energy consumption in China was 10 500 PJ; nearly 1.6 times greater than rural household energy consumption (6700 PJ) ( figure 1(A)). Heat (21.1%), electricity (21%), and gasoline (20.5%) accounted for 63% of total urban household energy consumption. In contrast, raw coal accounted for 34% of the total rural household energy consumption (1300 PJ), representing 2.8 times the amount of coal consumed by urban households. Electricity was the second largest fuel type used by both rural and urban households, that accounted for 1100 PJ and 1400 PJ, respectively ( figure 1(B)). Figure 1(B) indicates that fuel type usage is more homogenous than urban households, with the top three fuel types representing 77% of the total rural household energy consumption, whereas in urban areas, the top three fuel types represented 63% of the fuel used. To examine regional differences between urban and rural dominated regions, we divide China into eight economic zones according to a definition developed by the Development Research Center of the State Council (Wu 2014) (table S2) . These eight zones comprise the Northern coastal, Eastern coastal, Southern coastal, Northeastern, Northwestern, and Southwestern economic zones, as well as two economic zones defined to cover the middle reaches of the Yellow River and the middle reaches of the Yangtze River. Higher household energy consumption is more located in the Northern coastal, Eastern costal, Southern coastal and middle reaches of the Yellow River economic zone (figures 2(A) and (B)). Rural household energy consumption is highest in Hebei, Guangdong, Shandong, Henan and Hunan, that were 448, 293, 265, 243 and 202 PJ, respectively. For urban household energy consumption, consumption is highest in Guangdong, Liaoning, Shandong, Hunan and Heilongjiang, 553, 472, 433, 392 and 386 PJ, respectively. Regrading energy intensity, rural households located in Hebei, Shanxi, Gansu, Guizhou and Chongqing, 3.3, 2.7, 2.7, 2.6 and 2.6 GJ/US$ had the highest energy intensities. Urban households in Xinjiang, Heilongjiang, Shanxi, Inner Mongolia and Qinghai, 2.6, 2.1, 1.6, 1.4 and 1.3 GJ/US$, respectively, on average, energy intensity of rural households are larger than urban households and the Northern coastal, Northeastern, Northwestern, Eastern coastal and Southern coastal economic zone had higher energy intensities of household consumption.   Figure 3 presents the average population weighted PM 2.5 concentration calculated by the GAINS model (the comparison between ambient PM 2.5 concentration from the GAINS and 941 observation data from observing sites is shown in figure S4). The average population weighted concentration of PM 2.5 was 37 µg m −3 in China in 2015, which has exceeded the national ambient air quality standard (NAAQ) (35 µg m −3 ) of China and World Health Organization (WHO) air quality guideline (10 µg m −3 ). PM 2.5 concentration was generally higher the Northern coastal economic zone (Jing-Jin-Ji area) with the highest average population weighted PM 2.5 concentration, middle reaches of the Yellow River economic zone, middle reaches of the Yangtze River economic zone, Eastern coastal economic zone, Southern coastal economic zone, Sichuan Basin area in the Southwestern economic zone and Tarim Basin in the Northwestern economic zone and especially in urban areas. The five provinces with the highest population weighted PM 2.5 concentration were Henan (68 µg m −3 ), Hebei (67 µg m −3 ), Beijing (66 µg m −3 ), Tianjin (65 µg m −3 ) and Shandong (58 µg m −3 ) (figure 3). 55% of the population is exposed to annual average PM 2.5 concentration of more than 35 µg m −3 in 2015.

Premature deaths of household consumption
The total premature mortality attributable to PM 2.5 concentration in China across five health endpoints of interest, COPD, IHD, LC, LRI and Stroke were 1540 000 (95% CI:1270 000-1789 000).
The age-and sex-specific premature deaths attributable to PM 2.5 pollution associated with household energy consumption to varies across rural and urban areas. Premature deaths attributable to PM 2.5 pollution are greater in rural than urban households ( figure 5(A)). Figure 5 shows that IHD was the largest health burden attributable to household sourced PM 2.5 exposure (between 37.5% and 37.8% premature deaths). The population aged over  80 year-old accounted for over half the total household energy consumption-PM 2.5 -related premature deaths (66.3%-66.5%), with the age category 25-29 year-old recording the lowest. Premature mortality is higher among the male population compared to the female population, ranging from 62.3% to 62.4% of household energy consumption PM 2.5 -related premature deaths ( figure 5(B)).
The number of premature deaths per 100 000 people had a positive correlation with PM 2.5 concentration, as shown in figure 6, with rural household energy consumption attributable to a larger number of premature deaths per 100 000 people compared to urban household. When the PM 2.5 concentration was less than 45 µg m −3 , premature deaths was less than 20 per 100 000. Thirteen provinces (Henan, Hebei, Beijing, Tianjin, Shandong, Anhui, Shanghai, Jiangsu, Hubei, Chongqing, Shaanxi, Shanxi and Sichuan) had PM 2.5 concentration over 45 µg m −3 , with associated premature deaths per 100 000 people, ranging from 65.9 (Chongqing) to 2.6 (Shanxi). Chongqing, Shanghai, Beijing, Hubei and Sichuan had the highest estimated premature deaths per 100 000 (figure 6). Figure 7 shows the provincial distribution of premature deaths per 100 000 people and premature deaths per 100 000 people by age and sex-specific health endpoints by rural and urban household energy consumption. The number of premature deaths associate with rural household energy consumption per 100 000 of the population were highest in Chongqing, Shanghai, Beijing, Hubei and Tianjin (65.9,31.7,28.8,22.1 and 21.5,respectively); while the number of premature deaths per 100 000 associated with urban energy consumption was highest in Chongqing, Shanghai,Beijing,Hubei and Sichuan (65.7,31,22.5,21.5 and 21,respectively). Although the highest levels of PM 2.5 concentration were in the Northern coastal economic zone of China, the highest numbers of premature deaths per 100 000 people were more located in the Southern area of China which include the Southern coastal, the middle reaches of the Yangtze River and Yellow River, Eastern coastal economic zones, and the Sichuan Basin area in the Southwestern economic zone. Provincial level premature deaths per 100 000 attributable to PM 2.5 pollution by age-and sex-specific health endpoint. Details of provincial health burden from household consumption can be seen in figure S5.

Avoided premature deaths and economic benefits from reduced direct coal consumption
If rural and urban households in China substituted electricity for coal and biomass, the model suggests that this could result in a significant reduction in deaths: such substitution could save 37 400 (95% CI: 31 800-41 800) deaths attributable to rural household consumption and 6900 (95% CI; 6100-7900) lives attributable to urban household consumption, meaning on average 3 lives per 100 000 people can be saved in 2015. This corresponds to 28% and 6% of the total premature deaths caused by rural and urban household consumption in 2015. Economically, such a shift would result in US$ 2.5 billion (95% CI: 2.1-2.7) economic benefits for rural households, equaling 0.09% (95% CI: 0.08%-0.1%) of GDP of the rural region of China in 2015, and it could create 522 (95% CI: 468-594) million US$ economic benefits by urban household consumption, equaling 0.006% (0.005%-0.007%) of GDP of the urban region of China in 2015.

Discussion
A number of recent studies have examined the health burden attributable to PM 2.5 exposure in China (Song et al 2017, Maji et al 2018, Liu et al 2021, and shown the scale of the challenge. This study confirms the scope and scale of the challenge and focuses specifically on the direct role of household energy consumption specifically. The contribution here builds on knowledge of household energy use , Yun et al 2020 to implement an integrated assessment approach connecting energy, emission, air pollution and health outcomes. Hence, we investigate the adverse impacts of household energy consumption by various fuel types across rural and urban areas on PM 2.5 related age-and sex-specific premature deaths at provincial levels for 2015. This provides new insights by estimating the premature death associated with each of the 15 fuel types used by households in China across different age and gender groups at the national/regional/provincial levels for China. We further apply a scenario analysis to calculate the economic benefits from switching solid fuels into electricity. Our estimates found that in 2015, 17% of national premature deaths could be attributed to outdoor PM 2.5 from residential energy sector. Although urban households consumed nearly 1.6 times energy than rural households, premature deaths attributable to PM 2.5 exposure from household energy was 1.1 times higher from rural household consumption compared to urban households due to rural households use of solid fuel products. Regarding urban-rural differences, these findings are consistent with a similar study by Zhao et al (2019) using data from 2012. However, Zhao et al (2019) reported higher estimates for premature deaths for rural areas compared to our study (18% vs 8.7%). Direct energy consumption was higher in 2012 than 2015, but the difference may also be underpinned by to cleaner energy sources in rural. In 2012, raw coal energy usage corresponded to 50% and 1.3% of the total energy consumption in rural and urban households, whereas in 2015, the proportion of raw coal proportion had decreased to 46% and 0.9% for respectively rural and urban households (NBSC 2016).
Analysis at the regional level incorporating differences between urban and rural areas and age-sex specific mortality rates by five health outcomes, finds that between 37.5% and 37.8% premature deaths attributable to household energy consumption were due to IHD. Importantly, the analysis found that the distribution of household energy consumptionrelated premature deaths is not just a product of energy consumption and outdoor PM 2.5 concentration, but also population density and demographic structure (figures S6 and S7). The population aged over 80 year-old accounts for over half the total household energy consumption-PM 2.5 -related premature deaths (66.3%-66.5%), with the age category 25-29 year-old recording the lowest. Premature mortality was higher among the male population compared to the female population, ranging from 62.3% to 62.4% of household energy consumption PM 2.5 -related premature deaths. Household energy consumption-related premature deaths are shown to be highest in the Southern area of China, explained by a combination of high population density and an aging population. For example, Chongqing has the highest number of premature deaths as well as premature deaths per 100 000 and this was mainly attributable to its aging population (highest among the 30 provinces, figure S7), population density (listed as 11/30) and PM 2.5 concentration (listed as 10/30).
The scenario analysis finds that if coal and biomass had been replaced with electricity in both urban and rural households, 28% (rural) and 6% (urban) premature deaths would have been avoided. Previous research by Zhao et al (2018) similarly found that if solid fuels by Chinese household had been replaced with clean fuels, it could have saved 33% of the PM 2.5 -induced mortality in 2015. With regard to the lower estimates presented here, Zhao et al (2018) also calculated the avoid deaths from indoor air pollution. This is equivalent to US$ 2.5 billion (95% CI: 2.1-2.7) economic benefits for rural households and US$ 522 million (95% CI: 468-594) for urban household consumption. The estimates presented here are also lower than those by Yun et al (2020), who found that the residential sector contributed to 71% of the indoor PM 2.5 concentrations and 67% of PM 2.5 -induced premature deaths in 2014 in China. However, once again our model may underestimate pre-mature deaths as we do not account for indoor household energy consumption.
This research suggests that households in Beijing would receive the largest economic benefits from cleaner air. However, it is important to note that the value of statistical life approach inevitably gives disproportionate weight to wealthier regions because of the innate characteristics of the method. Larger potential economic benefits of cleaner air to regions with high concentrations of wealth and income remains a significant challenge in policy and decisionmaking, not least in the context of just and fair decarbonization transitions (Friel et al 2008). This study illustrates the large positive health and economic impacts that would be obtained from a shift to cleaner energy types within households in China, particularly regarding rural households.
From a policy perspective, this analysis suggests that mitigation measures such as promoting cleaner household fuel, through the subsidization of modern stoves within rural household will have large health and economic impacts, particularly in rural China. Li et al (2019) found that switching from solid fuels into carbonized fuels (higher thermal efficiencies and lower pollutant emissions) can generate environmental benefits for household residents. However, the health impacts of such a switch need to be more fully explored. For example, a recent clinical trial in the USA found that cookstoves emitting lower PM 2.5 emissions still had a negative impact on cardiac health (Cole-Hunter et al 2021). Irrespective of measures, the role of local municipalities will be crucial for the promotion and operationalization of new technologies, especially communities in poorer areas with resources are fewer higher baseline health risks than those in richer areas (Liu et al 2021).
There are two major limitations in this research which can lead to further research. First, our estimates do not account for indoor household energy consumption. Indeed research by Yun et al (2020) found that the residential sector contributed to 71% of the indoor PM 2.5 concentrations and 67% of PM 2.5induced premature deaths in 2014. As such this analysis is (a) a lower bound estimate of total premature deaths from direct household energy consumption and (b) likely underestimates the impact on women as they may well have higher exposure to indoor air pollution due to their longer duration indoors in residential settings, as suggested by Hashim and Boffetta (2014). Secondly, we do not consider fuel sources of power producing in our scenario analysis as electricity supplies are mostly from coal power plants in China (Hubacek et al 2009), hence the health cobenefits from fuel switching and decarbonizing in our research may be even more significant than portrayed.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).

Contributors
C L contributed to this study concept and design, did data collection, data analysis, data interpretation, visualization and wrote the draft of the manuscript. S Z supervised research, methodology and provided revision. Z L, W N A and K M supervised the research and revision. C T, T S, H Y and J G participated in data input into models. Y L participated in visualization and data collection. All authors approved the final version.