Accelerating the energy transition towards photovoltaic and wind in China

China’s goal to achieve carbon (C) neutrality by 2060 requires scaling up photovoltaic (PV) and wind power from 1 to 10–15 PWh year−1 (refs. 1–5). Following the historical rates of renewable installation1, a recent high-resolution energy-system model6 and forecasts based on China’s 14th Five-year Energy Development (CFED)7, however, only indicate that the capacity will reach 5–9.5 PWh year−1 by 2060. Here we show that, by individually optimizing the deployment of 3,844 new utility-scale PV and wind power plants coordinated with ultra-high-voltage (UHV) transmission and energy storage and accounting for power-load flexibility and learning dynamics, the capacity of PV and wind power can be increased from 9 PWh year−1 (corresponding to the CFED path) to 15 PWh year−1, accompanied by a reduction in the average abatement cost from US$97 to US$6 per tonne of carbon dioxide (tCO2). To achieve this, annualized investment in PV and wind power should ramp up from US$77 billion in 2020 (current level) to US$127 billion in the 2020s and further to US$426 billion year−1 in the 2050s. The large-scale deployment of PV and wind power increases income for residents in the poorest regions as co-benefits. Our results highlight the importance of upgrading power systems by building energy storage, expanding transmission capacity and adjusting power load at the demand side to reduce the economic cost of deploying PV and wind power to achieve carbon neutrality in China.


S1. Geospatial data used in the optimisation model
We compiled the data of land-cover at a spatial resolution of 0.005°0.005°from the Moderate Resolution Imaging Spectroradiometer (MODIS) Land Cover Type 1 (MCD12Q1) data set (United States Geological Survey, 2014).All land pixels were categorized into forests, shrublands, savanna, grassland, wetland, croplands, urban and built-up lands, natural vegetation mosaics, snow and ice, desert and water bodies.The suitability of installation of PV panels or onshore wind turbines was defined based on land-cover in each pixel (Supplementary Table 3).
To estimate the area of pixels suitable for installing PV panels and onshore-wind turbines for power generation, we compiled the area of terrestrial ecological reserve at a spatial resolution of 0.001°0.001°from the Resource and Environment Science and Data Center (Resource and Environment Science and Data Center, 2020), the slope of ground at a spatial resolution of 0.001°0.001°from the Shuttle Radar Topography Mission (SRTM) global enhanced slope data set (United States Geological Survey, 2015), and the zero-plane displacement height and the surface roughness at a spatial resolution of 0.5°0.625°from the NASA's Goddard Earth Observing System Model, version 5 (GEOS-5) Forward Processing (FP) database (Global Modeling and Assimilation Office, 2021).
To estimate the area of pixels suitable for installing offshore-wind turbines for power generation, we compiled the area of China's territorial sea from the Maritime Boundaries Geodatabase We compiled the hourly solar radiation and surface air temperature during 2012-2018 at a spatial resolution of 0.25° in latitude and 0.31° in longitude from the NASA's Goddard Earth Observing System Model, version 5 (GEOS-5) Forward Processing (FP) database (Global Modeling and Assimilation Office, 2021), which were re-projected on the map at a spatial resolution of 0. 0083° in latitude and 0.033° in longitude by assuming the homogeneity within the pixel of 0.25°0.31°.We compiled the hourly ground friction velocity of wind speed at a spatial resolution of 0.5° in latitude and 0.625° in longitude during 2012-2018 from the MERRA-2 data set (Gelaro et al., 2017), which were re-projected on the map at a spatial resolution of 0. 0083° in latitude and 0.033° in longitude by assuming homogeneity within the pixel of 0.5°0.625°.
We compiled data of energy consumption in the power, residential, industry, transportation and other sectors by province in 2019 from the Chinese Energy Statistical Yearbook 2020 (National Bureau of Statistics of China, 2020).The amounts of energy consumption are projected by sector and by province from 2021 to 2060 based on the energy consumption in 2020 and the rate of growth of energy demand in China by year according to the projection by the International Energy Agency (International Energy Agency, 2021).We disaggregated the energy consumptions from 31 provinces over the mainland to 2,373 counties using the spatial distribution of GDP at a spatial resolution of 0.01°0.01° in 2019 from the Resource and Environment Science and Data Center (REDC) (Xu, 2017) due to a lack of data for Hongkong, Macao and Taiwan.Geospatial data used in the optimisation model are summarized in Supplementary Table 2.

S2. Calculation of wind speed at a hub height
Wind power is subject to large seasonal and diurnal variabilities.To deliver a representative estimate of wind power resources for 2021-2060, we estimated the hourly wind speed at a hub height of 100 meters above ground in the period for 2012-2018 using a log-law function (Rinne et al., 2018): where V is wind speed at a height (h) of 100 meters above ground and k is the Von Karman constant (0.41).The hourly friction velocity of wind speed (u * ) was compiled from the MERRA-2 data set at a spatial resolution of 0.5°0.625°(Gelaro et al., 2017), while the zeroplane displacement height (d) and the surface roughness (z0) were compiled from the NASA's GEOS-5 FP database at a spatial resolution of 0.5°0.625°(Global Modeling and Assimilation Office, 2021).

S3. Power generation by photovoltaic-power plants
We identified the area of pixels suitable for utility-scale power generation by photovoltaic (PV) plants with a capacity of >10 MW using solar energy based on the type of land-cover, ground slope, solar radiation, surface air temeperature and the proven area of terrestrial ecological reserve.We defined the suitability factor for suitable lands varying from 10 to 15% based on the land-cover, while we excluded the area of natural reserve with ecological functions and the pixels with ground slope >5%, solar radiation <4.2 hour per d and surface atmospheric temperature <0℃ (Supplementary Table 3).
Electricity generation by PV-power plants is influenced by solar radiation, the effective area of PV panels, the geographic location of PV panel, surface atmospheric temperature, the shade between PV panels and the efficiency of energy conversion (Chen et al., 2021).We estimated the capacity potential of PV panel (PPV) (Masters, 2013): where Spanel is the area of pixels installed with PV panels, Ω is the ratio of effective panel area to the area of pixels installing with PV panels, and SR is the standard power capacity of PV panels (161.9Watt per m 2 ) (Masters, 2013).
We estimated the hourly power generation (WPV) by a PV-power plant as a function of capacity potential (PPV) (Masters, 2013;Chen et al., 2019): where Ipanel is solar radiation captured by PV panels, I0 is intercepted radiation of PV panels (1,000 Watt per m 2 ) under a standard test condition, γshade is the shade coefficient, γtemp is the temperature coefficient, and γloss is electricity loss from power generation to grid connection (19.44%).We estimated Ipanel, γshade and γtemp using the recommended methods (Masters, 2013;Chen et al., 2019), which are described below.
First, we estimated solar radiation (Ipanel) captured by PV panels as the sum of direct (Idirect), diffuse (Idiff) and reflected radiation (Iref) (Masters, 2013): where ρs is the surface albedo (0.2) (Chen et al., 2019).The hourly solar direct (Rdirect), diffuse (Rdiff) and total (Rtotal) radiation were compiled from the NASA's GEOS-5 FP database at a spatial resolution of 0.25°0.31°(Global Modeling and Assimilation Office, 2021).It should be noted that there is an ongoing debate on how the surface solar radiation has changed in different regions (Wild et al., 2015) and what has caused the regional change in surface solar radiation (i.e.effects of aerosols or cloud cover) (Imamovic et al., 2016).However, it is also known that historical changes in surface solar radiation are not well reproduced by climate models, thus limiting our ability for predicting future changes in surface solar radiation (Moseid et al., 2020).
Second, we estimated the shade coefficient (γshade) to consider the impact of shading between PV panels on the efficiency of solar energy capture (Chen et al., 2019): where γshade is the proportion of the unshaded area relative to the total area of PV panels.
Third, we estimated the temperature coefficient (γtemp) be considering the negative effect of extremely high temperatures on the efficiency of converting solar energy into electricity (Kaldellis et al., 2014;Kawajiri et al., 2011): wher Tatm is the hourly atmospheric surface temperature at 2 meters above ground, Tcell is the normal cell operating temperature (44 ℃) and σT is the temperature coefficient (-0.41% per ℃) (Chen et al., 2019).

S4. Power generation by onshore wind-power plants
We identified the area of pixels suitable for utility-scale power generation by onshore-wind power plants with a capacity of >10 MW based on the type of land-cover, ground slope, altitude of sites above sea-level and area of terrestrial ecological reserve (Supplementary Table 3).In our assumption, onshore wind-power plants are installed with the General Electric wind turbine at a maximal capacity of 2.5 MW at a hub height of 100 meters above ground to convert air kinetic energy to electricity.The power generation curve as a function of wind speed at a height of 100 meters for the General Electric model in the Wind-turbines database (https://en.windturbine-models.com/turbines) was adopted to estimate the capacity factor of onshore windpower generation (Lu et al., 2020; Bauer and Matysik, 2021) (Supplementary Fig. 3).
Following a method in the literature (Lu et al., 2020), we estimated the onshore wind-power density (onshore) by installing wind turbines with an 8×8 rotor diameter to improve the efficiency of kinetic energy conversion with a low turbine-turbine interference on lands: 8D onshore ×8D onshore (14) where PWonshore is the maximal power of onshore-wind turbine (2.5 MW), and Donshore is the diameter of rotor for onshore-wind turbines (103 meters).
Based on the onshore-wind power density (onshore) and the area of pixels installed with onshore wind turbines (Sonshore), we estimated the capacity potential of onshore-wind power (Ponshore) (Masters, 2013): We estimated the hourly power generation by an onshore wind-power plant (Wonshore): W onshore =P onshore ×CF onshore ×U TI ×A RR (16) where CFonshore is the capacity factor of onshore-wind turbines calculated by the power generation function, UTI is the efficiency of energy conversion (0.95) (Rinne et al., 2018), and ARR is the array efficiency factor (0.9) (Rinne et al., 2018).

S5. Power generation by offshore wind-power plants
We identified the pixels suitable for utility-scale power generation by offshore-wind plants with a capacity of >10 MW using kinetic energy over the sea (Lu et al., 2020;Sherman et al., 2017).
We considered that offshore-wind plants will be constructed in the oceans with water depth <60 (Supplementary Fig. 3).
We estimated the power density (offshore) of offshore-wind turbines using a 7×7 rotor diameter to improve the efficiency of wind energy utilization with a low turbine-turbine interference on oceans (Lu et al., 2020;Sherman et al., 2017): where PWoffshore is the maximal power of offshore wind turbine (8 MW), and Doffshore is the diameter of rotor for offshore wind turbines (164 meters).

S6. Costs of ultra-high-voltage transmission of electricity
We considered that electricity will be transported among regions in a national grid network of ultra-high-voltage (UHV) transmission using 130 lines that have been projected in China's national grid development plans and 817 lines that have not been projected but are needed in our optimisation model before 2060 (Center for Security and Emerging Technology, 2021) (see the detailed information on these UHV lines in the Supplementary Spreadsheet S1).The direct current (DC) UHV is designed at a capacity of 8,000 MW for ±800 kV DC and 12,000 MW for ±1,100kV DC, respectively (Chen et al., 2021).In contrast, the UHV transmission capacity of alternating current (AC) is a function of the transmission distance estimated in a previous study (Chen et al., 2021), where the capacity of 1,000 kV AC decreases from 6,000 MW for 100 km to 3,000 MW for 3,000 km.Based on the capacity and distance of electricity transmission, we estimated the costs of transmission (Aϵ) (Chen et al., 2021): where l is a transmission line, μline is the line cost per kilometer ($732,220, $800,383 and $670,785 per km for ±800 kV DC, ±1,100kV DC and 1,000 kV AC, respectively) (Electric Power Planning and Engineering Institute, 2011; Electric Power Planning and Engineering Institute, 2020), Dl is the length of a transmission line, Θhl is the hourly electricity carried by a transmission line, PUHV is the capacity of a transmission line (8,000 MW or 12,000 MW for DC and 3,000 to 6,000 MW for AC), and μsub is unit costs of converters for DC lines or substations for AC lines ($82, $92 and $41 per kW for ±800 kV DC, ±1,100kV DC and 1,000 kV AC, respectively) (Electric Power Planning and Engineering Institute, 2020).

S7. Costs of energy storage by hydro pump or chemical batteries
By considering that chemical batteries can be charged and discharged for 6,000 times over a lifetime of 15 years (Chen et al., 2021) and that pumped-hydro storage can be charged and discharged for 1 time per day over a lifetime of 50 years (Cole and Frazier, 2019), we estimated the costs of energy storage (Gϵ): where x is a power plant, h is an hour, q is a region, nq is the number of power plants, Px is the capacity potential of storage systems (maximal hourly electricity in storage), Λh is the hourly electricity in storage, μpower is unit costs of power capacity ($1,200 per kW for pumped-hydro storage in 2020-2060; $595, $374, $327, $280 and $234 per kW for chemical battery storage we sought the optimal option of energy storage to achieve the lower LCOE for each power plant.

S8. Intertemporal dynamics of learning
We adopted the formulation of learning by doing (Mcdonald and Schrattenholzer, 2001) to estimate the ratio of the declined capital costs (ξx) by accumulating low-carbon investments: where x is a new PV or wind-power plant, ς is a PV or wind-power plant build before x, tx is the time of building plant x, tς is the time of building plant ς, Pς is the power capacity of plant ς, q is a region, nq is the number of power plants in a region, P0 is the total capacity of PV or wind power built by 2020 (253,000 MW, 272,010 MW and 8,990 MW for PV, onshore wind and offshore wind, respectively) (China Energy Storage Network, 2021), and rLR is the rate of learning.We derived the average of rates of learning (32.4% for PV panels and 11.78% for wind turbines, 18% for inverters, mounting materials, secondary equipment, installation work and administration and grid connection, and 18% for power transmission) from measurements in China and considered their uncertainty ranges in our Monte Carlo simulations (Supplementary Table 1).

S9. Estimation of the Gini coefficient for income inequality
We estimated the Gini coefficient to represent income inequality (Yitzhaki, 1979) by dividing all population into 2,002 groups in the order of per capita income.We estimated the income Gini coefficient in China following a simple formula (Mi et al., 2020): where i is a population group in the country, ni is the number of groups in the country (2,002), g is a group with per capita income lower than that in group i, Γg or Γi is the ratio of income in group g or i to the total income of the population in the country, and Ψi is the ratio of population in group i to the total population in the country.We estimated Γg, Γi and Ψi in the country based on the income distribution in 2,373 counties in China in 2060 by varying the price of carbon (ϛ) from $0 to $100 per tCO2.
We Under a prescribed carbon price (ϛ), we sought for the power plants with the marginal abatement costs lower than this carbon price, returning the total revenue (Rϵ) of power generation after building a new PV or wind-power plant (ϵ): where ϱ is the price of coal, oil or gas substituted by PV or wind power, Fϵ, Eϵ and LCOEϵ are abated CO2 emissions, power generation and LCOE after building plant ϵ.We then obtained the revenue in each county by building PV and wind-power plants as: where x is a county building the power plant ϵ, and nx is the number of PV and wind-power plants in this county.
We filtered pixels in urban area when building new PV or wind-power plants, so we allocated the revenue among the rural population in China as a part of poverty alleviation project in the country.By considering payments due to the increase in the costs of power generation after levying a carbon tax on fossil fuels, we estimated the per capita income of each population group in rural and urban areas: where x is a county (x=1 to 2,373), i is a group in the rural or urban population (i=1 to 1,001), 1001 i=1 (30) where pi,x is the number of population in a rural group, qi,x is the number of population in an urban group, INRi,x,0 is the per capita income for rural group x without building PV and wind-

405
Table S1.Learning rates for photovoltaic (PV) and wind power in the literature.

Land-cover
The land-cover data in 2019 at a spatial resolution of 0.005°0.005°were compiled from the Moderate Resolution Imaging Spectroradiometer (MODIS) Land Cover Type 1 (MCD12Q1) data set (United States Geological Survey, 2014), which were used to estimate the suitability factor and the terrestrial carbon sink (https://lpdaac.usgs.gov/products/mcd12q1v006/).

Solar radiation
The geospatial data of hourly direct and diffuse solar radiation during 2012-2018 at a spatial resolution of 0.25° in latitude and 0.31° in longitude were compiled from NASA's Goddard Earth Observing System Model, version 5 (GEOS-5) Forward Processing (FP) data set (Global Modeling and Assimilation Office, 2021), which were used to identify the pixels suitable for installing PV panels and estimate the power generation potential of PV energy (https://portal.nccs.nasa.gov/cgi-lats4d/opendap.cgi?&path=GEOS-5/fp/0.25_deg/assim).
Wind speed at a hub height of 100 meters above ground (V) The hourly wind speed at a height of 100 meters above ground at a spatial resolution of 0.5° in latitude and 0.625° in longitude during 2012-2018 were calculated based on the hourly friction velocity, the displacement length and the roughness length from the Modern-Era Retrospective analysis for Research and Applications, version 2 (MERRA-2) data set (Gelaro et al., 2017) (https://portal.nccs.nasa.gov/cgi-lats4d/opendap.cgi?&path=GEOS-5/fp/0.25_deg/assim).

Air temperature at 2 meters (Tatm)
The geospatial data of hourly air temperature at 2 meters above ground at a spatial resolution of 0.25° in latitude and 0.31° in longitude during 2012-2018 were compiled from the NASA's Goddard Earth Observing System Model, version 5 (GEOS-5) Forward Processing (FP) data set (Global Modeling and Assimilation Office, 2021), which were adopted to identify the pixels suitable for installing PV panels and estimate the power generation potential of PV energy (https://portal.nccs.nasa.gov/cgi-lats4d/opendap.cgi?&path=GEOS-5/fp/0.25_deg/assim).

Ground slope
Ground slope data at a spatial resolution of 0.001°0.001°observed in 2000 were compiled from the Shuttle Radar Topography Mission (SRTM) global enhanced slope data set (United States Geological Survey, 2015), which were used to identify the pixels suitable for installing PV panels and onshore wind turbines (https://lpdaac.usgs.gov/products/srtmgl1v003/)

Mask of Territorial Sea Area (TSA)
Territorial Sea Area in 2018 were compiled from the Maritime Boundaries Geodatabase (Flanders Marine Institute, 2018) to identify the pixels suitable for installing offshore wind turbines (http://www.vliz.be/en/imis?dasid=5465&doiid=312).

Mask of natural reserves
Mask of terrestrial ecological reserve in 2020 at a spatial resolution of 0.001°0.001°were derived from the Resource and Environment Science and Data Center (Resource and Environment Science and Data Center, 2020) (http://www.resdc.cn/data.aspx?DATAID=137), and the mask of marine ecological reserve in 2021 were derived from the National Marine Data and Information Service (UN Environment Programme World Conservation Monitoring Centre, 2021; Resource and Environment Science and Data Center, 2020), which were adopted to identify the pixels suitable for installing PV panels, onshore wind turbines and offshore wind turbines (https://www.protectedplanet.net/country/CHN).

Water depth (DP)
The geospatial data of water depth at a spatial resolution of 0.001°0.001°observed in 2000 were compiled from the Radar Topography Mission (SRTM) Global Enhanced Slope (GES) data set (United States Geological Survey, 2015) to identify the pixels suitable for installing offshore wind turbines (https://lpdaac.usgs.gov/products/srtmgl1v003/).
Friction velocity (u*), zero-plane displacement height (d), and surface roughness (z0) The geospatial data of hourly friction velocity, zero-plane displacement height and the surface roughness during 2012-2018 at a spatial resolution of 0.5°0.625°were compiled from the NASA's Goddard Earth Observing System Model, version 5 (GEOS-5) Forward Processing (FP) database (Global Modeling and Assimilation Office, 2021), which were used to estimate the wind speed at the hub height of 100 meters above ground (https://portal.nccs.nasa.gov/cgi-lats4d/opendap.cgi?&path=GEOS-5/fp/0.25_deg/assim).Capacity factor of wind turbines (CFonshore and CFoffshore) The capacity factor of onshore and offshore wind turbines as a function of the wind speed was predicted using the wind turbine models available at https://en.wind-turbine-models.com/turbines (Bauer and Matysik, 2021).

Terrestrial carbon sink (γx)
The geospatial data of terrestrial carbon sink were obtained by performing a disaggregation of the total net terrestrial carbon sink uptake in China (1.87 Gt CO2 y -1 ) inferred from measurements of atmospheric CO2 gradients from 2006-2009 (Jiang et al., 2016)

Spatial distribution of GDP
The spatial distribution of GDP in 2015 at a spatial resolution of 0.01°0.01°were compiled from the Resource and Environment Science and Data Center (Xu, 2017) (https://www.resdc.cn/data.aspx?DATAID=252).

Annual mean temperature change
The change in annual mean surface air temperature during 2021-2060 for China were derived from simulations in the SSP1-2.6 scenario in an Earth system model (Gasser, 2017).

Traffic flow data
The profile of traffic flow was derived from observations in each street per 5 minutes in 2018 in Shenzhen as a megacity in China (Shenzhen Municipal Government data open platform, 2018)

Historical
The typical power load profiles by month in each province in 2018

Income distribution for urban and rural populations
The residents' income distribution for urban and rural populations were compiled from the national socioeconomic survey in 2015 (Zhang et al., 2016).
Table S3.Indicators used to filter the pixels suitable for installing PV panels or wind turbines.

(
Flanders Marine Institute, 2018), depth of water from the Radar Topography Mission (SRTM) Global Enhanced Slope Database (United States Geological Survey, 2015), routes of shipping from the Modern-Era Retrospective analysis for Research and Applications, version 2 (MERRA-2) database (Gelaro et al., 2017), and the area of marine ecological reserve from the National Marine Data and Information Service (UN Environment Programme World Conservation Monitoring Centre, 2021; Resource and Environment Science and Data Center, 2020).
in 2020, 2030, 2040, 2050 and 2060, respectively) (Hiesl et al., 2020; Cole and Frazier, 2019), μthrou is unit costs of throughput in the storage systems ($100 per kWh for pumped-hydro storage in 2020-2060; $345, $198, $174, $149 and $124 per kWh for lithium battery storage in 2020, 2030, 2040, 2050 and 2060, respectively) (Chen et al., 2021; Cole and Frazier, 2019), μopera is unit operational costs in charging and discharging ($0.0015 per kWh) (Zhang et al., 2016), Nc is the number of charging and discharging (365 and 400 for pumped-hydro storage and chemical battery storage, respectively) (Chen et al., 2021), ε is the ratio of energy after charging and discharging (85%) (Cole and Frazier, 2019), t is the lifetime of storage (50 and 15 years for pumped-hydro storage and chemical batteries, respectively) (Chen et al., 2021; Cole and Frazier, 2019), and r is the discounting rate (5% per y) (Duan et al., 2021).By comparing LCOE for each power plant using mechanical (pumped hydro) or chemical (chemical batteries) storage, estimated Γg, Γi and Ψi by considering the impact of finances embodied in the flow of PV and wind power based on income distribution of population in 2060.First, we estimated the income distribution of population by county based on the residents' income distribution for urban and rural population according to a national socioeconomic survey in 2015(Zhang et al., 2016).According to this income distribution, we estimated the average per capita income of urban and rural population by county and thus the frequency distribution of per capita income among urban and rural populations.To estimate the income distribution of population by county, we divided the urban or rural population into 1,001 groups.This method generates a total of 2,002 groups for urban and rural populations in each county.Second, we compiled the per capita disposable income among urban and rural population in 2,373 counties in China in 2015-2019 from the Provincial Statistical Yearbook (National Bureau of Statistics of China, 2020), and predicted per capita disposable income in 2060 based on the projected growth rate of income by province in China during 2020-2060 (National Bureau of Statistics of China, 2020).We assumed that the growth rate of per capita income is the same for all groups in a county due to the lack of data.We calibrated the growth rate of per capita income in each income group by county during 2015-2060 to ensure that the predicted per capita income as an average for each county in 2060 is equal to the projection in 2060 by the Provincial Statistical Yearbook (National Bureau of Statistics of China, 2020).This calculation returns the distribution of per capita income among rural and urban population by county in 2060.

Fig. S1 .Fig. S2 .Fig. S4 .
Fig. S1.Impacts of the discounting rate and lifetime of power plants on the average using the geospatial data of terrestrial carbon sink at a spatial resolution of 4° in latitude and 5° in longitude from 2010-2016 (Wang et al., 2020) as a proxy.Energy consumption by province Consumption of primary energy and electricity by province in the power, residential, industrial, transportation and other sectors in 2019 were compiled from the Chinese Energy Statistical Yearbook 2020 (National Bureau of Statistics of China, 2020).
with the emission rate of SO2 above 10 -11 kg m -2 s -1 (Mi et al., 2020)a)a income for a rural group, INUi,x is the per capita income for a urban RURx is rural population in county x, θfossil is CO2-emission factor of fossil fuels (0.84 kg CO2 per kWh for coal, 0.72 kg CO2 per kWh for oil and 0.46 kg CO2 per kWh for gas, respectively)(Liu et al., 2015a), mx is per capita energy consumption in county x, ϛ is the carbon price, and ui,x or vi,x is a factor to change the ratio of per capita energy consumption in a rural or urban group relative to the per capita energy consumption in this county.By assuming that the per capita income was proportional to per capita energy consumption in each county(Mi et al., 2020), we derived the per capita energy consumption as a function of per capita income: group, INRi,x,0 is the per capita income for rural group x without building PV and wind-power plants, INUi,x,0 is the per capita income for a urban group without building PV and wind-power plants, Tx is the revenue of building new PV and wind-power plants in this county,

Table S2 . Input data for the optimisation model. 408
(Duan, 2021)tion by nuclear, hydro, hydrogen and bioenergy in 2019 was compiled by province from National Statistical Yearbook of Energy (National Bureau of Statistics of China, 2020).The power generation by oil, gas, bioenergy, nuclear and hydropower were derived as the average of three IAMs (SWITCH, IPAC and GCAM_TU) in the "1.5℃-limiting" scenario for 2021-2060 from a mutli-model study(Duan, 2021).

Table S9 .
Soil carbon content by zone in China.The data are compiled from a national field 425 survey (Lai et al., 2016).

Table S10 . Projection of hourly power demand in different sectors by 2060.
(Gasser et al., 2017) during 2021-2060 was predicted based on the projected increase in power demand in the residential sector under an electrification rate of 58% by 2060 (International Energy Agency, 2021), the hourly gridded temperature in 2020 compiled from the Goddard Earth Observing System Model (https://gmao.gsfc.nasa.gov/GMAO_products/NRT_products.php) and the change in annual mean temperature during 2021-2060 over the China region under the SSP1-2.6 scenario simulated by an Earth system model(Gasser et al., 2017).TransportationThe hourly power load during 2021-2060 was predicted based on the projected increase in power demand in the transportation sector under an electrification rate of 58% (International Energy Agency, 2021), the hourly traffic flow data in Shenzhen city in 2018, the hourly gridded temperature in 2020 compiled from the Goddard Earth Observing System Model (https://gmao.gsfc.nasa.gov/GMAO_products/NRT_products.php) and the change in annual mean temperature during 2021-2060 over the China region under the SSP1-2.6 scenario simulated by an Earth system model(Gasser et al., 2017).AgricultureThe hourly power load during 2021-2060 is simulated endogenously by decade when the hourly power load profile is adjusted to match the hourly power generation by PV and wind power when meeting different targets of CO2 emission abatements.To obtain the hourly power load before this optimisation, the historical hourly power load profile from the provincial electrical grids in 2018 (National Development and Reform Commission, 2019) was scaled up by the projected increase in power demand during 2021-2060 under an electrification rate of 58% (International Energy Agency, 2021).