Global CO2 uptake of cement in 1930-2019

Because of the alkaline nature and high calcium content of cements in general, they serve as a CO2 absorbing 15 agent through carbonation processes, resembling silicate weathering in nature. This carbon uptake capacity of cements could abate some of the CO2 emitted during their production. Given the scale of cement production worldwide (4.10 Gt in 2019), a life-cycle assessment is necessary in determining the actual net carbon impacts of this industry. We adopted a comprehensive analytical model to estimate the amount of CO2 that had been absorbed from 1930 to 2019 in four types of cement materials including concrete, mortar, construction waste and cement kiln dust (CKD). Besides, the process CO2 emission during the 20 same period based on the same datasets was also estimated. The results show that 21.12 Gt CO2 (18.12-24.54 Gt CO2, 95% CI) had been absorbed in the cements produced from 1930 to 2019, with the 2019 annual figure mounting up to 0.90 Gt CO2 yr (0.76-1.07 Gt CO2, 95% CI). The cumulative uptake is equivalent to approx. 52% of the process emission, based on our estimation. In particular, China’s dominant position in cement production/consumption in recent decades also gives rise to its uptake being the greatest with a cumulative sink of 6.21 Gt CO2 (4.59-8.32 Gt CO2, 95% CI) since 1930. Among the four types 25 of cement materials, mortar is estimated to be the greatest contributor (approx. 58%) to the total uptake. Potentially, our cement emission and uptake estimation system can be updated annually and modified when necessary for future low-carbon transitions in the cement industry. All the data described in this study, including the Monte Carlo uncertainty analysis results, are accessible at https://doi.org/10.5281/zenodo.4064803 (Wang et al., 2020). https://doi.org/10.5194/essd-2020-275 O pe n A cc es s Earth System Science Data D icu ssio n s Preprint. Discussion started: 6 October 2020 c © Author(s) 2020. CC BY 4.0 License.


Introduction
According to the International Energy Agency (IEA) statistics, cement industry is the second largest industrial CO2 emitter with a share of 27% (2.2 Gt CO2/yr) in 2014 (IEA and WBCSD, 2018). Broadly, there are two direct sources of CO2 emission originating from cement production: 1. The thermal decomposition of limestone (CaCO3) in the process of producing clinker; 2. the energy required for the decomposition, largely provided by combustion of fossil fuels. For the latter, energy efficiency improvement and cement kiln technology advancement have gained noticeable progress in recent years (Shen et al., 2016;Xu 35 et al., 2014;Zhang et al., 2015). However, it has been widely estimated that the former so-called process emission constitutes most of the total direct emission (approx. 60%). Consequently, the targeted reduction in emission of cement industry for achieving Climate Action SDG 1 , which fully aligns with meeting the 'below 1.5℃' climate target (Rogelj et al., 2018), hinges upon reducing process emission. Unfortunately, the traditional standardised Ordinary Portland Cement (OPC), which has been the dominant type of cement used by humans so far, is of very high clinker contents historically i.e. high clinker-to-cement 40 ratio (herein referred as clinker ratio). Both Griffin et al. (1989) and Boden et al. (1995) reported the emission factor (EF) to be around 0.5 t CO2/t cement then, which suggested an implicit clinker ratio >95%. On the other hand, since OPC clinkers are CaO-rich, a high clinker ratio would also increase the CO2 absorption capabilities (by carbonation) of cements. The universal carbonation mechanisms that are responsible for the carbon uptake of cements can be attributed to their hydroxide(s) and silicate(s) constitutes, as described by Eq. (R1) and (R2) Pan et al. (2020) recently studied the emission reduction potential from producing cement mortar and concrete blocks by mixing in high level of alkaline blending (e.g. blast furnace slag, fly ash and mine tailings) and discovered a yearly multi-gigatonne potential of CO2 abatement. Therefore, reducing clinker ratio is still the key to lower the emission level of cement 50 industry while the projected demand for cement is going to increase by 1.1~1.2 times by the end of 2050 (IEA and WBCSD, 2018). Andrew (2018) updated the global cement industry emission inventory 2 recently by using various data sources for different countries and time periods. The insufficient accounting for the geographically and temporally varying clinker ratio, as was embedded in prior estimation methods adopted by Carbon Dioxide Information Analysis Centre (CDIAC) (Boden et 55 al., 2017), was considered and corrected for. On the other hand, in our previous study on the uptake (Xi et al., 2016), clinker ratio values from historical literature, including IPCC (2006) recommended default value of 0.75 (as the lower bound), were 3 used in our model for estimating the uptake as well as the uncertainty analysis by Monte Carlo method. Therefore, updating the results by applying more realistic clinker ratio data is necessary, especially for China where multiple surveys and reports have uncovered the strikingly lower-than-average clinker ratios post-1990. 60 In this study, we re-estimated the amount of CO2 uptake by cements produced from 1930 to 2019, including those used in concrete and mortar as well as those 'lost' as construction waste and kiln dust. We updated the clinker ratio/production data after 1990 for China and treated India as a separate region. We estimated that 21.12 Gt CO2 (18.12-24.54 Gt CO2, 95% CI) had been absorbed and sequestered in cements that had been produced between 1930 and 2019, which effectively abated 52% of the corresponding process emission. The annual uptake in 2019 alone reached a staggering 0.90 Gt CO2 yr -1 (0.76-1.07 Gt CO2 yr -1 ). Using this consistent framework and model, we could include regularly updated annual estimates of cement carbon uptake into annual assessments of the global carbon project (GCP) (Friedlingstein et al., 2019) as an important anthropogenic carbon sink, which has not been thoroughly assessed or documented.

Cement/clinker production data resources and treatment
Global cement production data have been estimated by United States Geological Survey (USGS) since 1930s. In our previous study (Xi et al., 2016), we used USGS production data explicitly as the only source for calculations of the uptake. In addition, the world was geographically divided into four primary countries/aggregated regions including China, the United States (US), Europe and Central Eurasia (including Russia) and Rest of the World (ROW). We noticed that, other than Russia 75 and Turkey, the country-specific European and Central Eurasian cement production data was not available yet from USGS after 2017. In this work, to keep the consistency with prior geographical division and data source, 2018 and 2019 cement production data were projected for the 'Europe and Central Eurasia' region. Specifically, the average ratio of the production in Russia and Turkey to the total production in 'Europe and Central Eurasia' from 2013 to 2017 was taken as the scaling factor, so that the total regional production for 2018 and 2019 can be projected assuming this proportion remained the same. For the 80 US, ROW and China (prior to 1990), we continued to use the cement production data since 1930 from USGS. The IPCC recommended clinker ratios were continually used for these aggregated regions without extra fine tuning to country-level data.
In terms of the updates on China, we first collected national cement production data for the period of 1990-2019 from  1990-1999, 2000-2014, and 2015-2019 periods were from published literature (Gao et al., 2017;Xu et al., 2012Xu et al., , 2014, China Cement Almanac (CCA, 2001(CCA, -2015, and public national data from Ministry of Information Technology (MIIT, 2019), respectively. As such, we also obtained the national clinker production for the 1930-  (Andrew, 2018(Andrew, , 2020 suggested that the average clinker ratio in India has been fluctuating in the past three decades. Therefore, we used the newly published year-by-year clinker ratio data for India for 1990-2019.

Estimating the process emission
Process CO2 emissions of the cement industry were estimated by multiplying regional clinker production by the derived process CO2 emission factors. Since the process CO2 emissions arise from chemical reactions involved in the production of 95 clinker, as carbonates (largely limestone, CaCO3) are decomposed into oxides (largely lime, CaO) and CO2 by the addition of heat, they can be estimated by the conservation of mass flow principle. The default value recommended by IPCC is 510 kg CO2 t -1 clinker (Hanle et al., 2006), without considering emissions originating from MgCO3. In this study, we first collected local survey data by kiln type from literature and applied them in the emission estimates. There are mainly five kiln types worldwide, including dry with preheater and precalciner, dry without preheater (long dry kiln), dry with preheater without 100 precalciner, wet/shaft kiln, and semi-wet/semi-dry.
For China, a nationwide sampling survey of 359 cement production lines across 22 provinces was conducted (Shen et al., 2016) and we adopted the process CO2 emission factor estimated from this local Chinese study. As a result, we applied the sample-averaged emission factors: 519.66 kg CO2 t -1 clinker for dry with preheater without precalciner, dry with preheater and precalciner, and dry without preheater (long dry) kilns, 499.83 kg CO2 t -1 clinker for semi-wet/semi-dry and wet/shaft kilns.

105
For other countries in the absence of detailed survey data, we adopted the emission factors that were collected and summarised in (Andrew, 2018), which integrated local emission information for key countries (e.g. India). We then obtained annual country-or regional-level production technology information from the World Business Council for Sustainable Development (WBCSD) and the Global Cement Directory 2019 (publicly named as the GCD-2019 dataset). While WBCSD collected technology-based clinker production information using a survey-based approach (IEA and WBCSD, 2018), the GCD-2019 110 dataset provides plant-level information of cement industries in service as of 2019 (for example, cement production capacity, physical address, number of kilns, and cement production technology etc.). We then cross-checked and integrated the 'start of operation year' information at plant level from the 'industryAbout' database (industryAbout, 2019) and various companies' websites. This information enabled us to infer the annual capacity-weighted production technology (i.e. kiln types) distributions for the 1930-2019 period. Finally, we used technology-weighted emission factors to calculate the regional average 115 emission factors, which were then used to estimate process CO2 emissions directly.
It is noted that in order to stay in line with the life-cycle CO2 uptake assessments of concrete structures, concrete construction waste and CKD in this study, in comparison to some previous studies (e.g. Andrew, 2018), our estimation framework for process CO2 emissions is relatively simple. Nevertheless, we integrated the global plant-level capacity and technology information into our estimates for the first time, to provide new perspectives on emission estimates. Besides, we also assessed 120 https://doi.org/10.5194/essd-2020-275 the uncertainties of such estimates using Montel Carlo method.

Life-cycle uptake assessments of concrete structures
Here, we adhere to the breakdown of concrete utilisation into 3 stages as before (Xi et al., 2016): 1. Service; 2. Demolition; 3. Secondary use. Therefore, the carbon uptake of concrete ( con C ) can be calculated as an aggregate of the three subcomponents: where tl l C , , td d C , and ts s C , are the uptake during service, demolition and secondary use stage, respectively. The life cycle was deemed to be 100 years in line with a historical study by Pade and Guimaraes (2007), considering the longest average life of buildings in Europe is merely 70 years (Pommer and Pade, 2005). During concretes' service life, they are used primarily to build various functional buildings, roads, utilities and other public works etc., hence exhibiting different sizes and geometrical 130 shapes in the environment. We adopt a simplified approach by considering a three-dimensional diffusion 'slab' model in which carbonation starts at the exterior side of the slab and gradually moves inwards: this is schematically shown in Figure 1.
According to Fick's 2 nd Law, which is used in the calculation 3 , the carbonation depth is proportional to the square root of the carbonation time (  and on the CaO content in the clinker ( f CaO clinker ). Additionally, in natural conditions, not all of the calcium in OPC would be associated with carbonation reactions due to its microstructural constraints (Lagerblad, 2005) hence the fraction of CaO that could be converted to CaCO3 (  ) should be considered, too, as follows:  Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License.

CO2 diffusion in the cement.
In order to estimate the carbon uptake at macroscopic scale with the data available, we made the following simplifications: 1. Assuming the diffusion front is equivalent to the carbonation front and the area behind the front is regarded fully carbonated; 2. Assuming the geometries of the cement parts resemble the slab shown in Figure 1 so that the exposed surface area ( i A ) can be calculated by the concrete volume in different structure categories and average thickness data. Further, since the carbonation 150 rate depends on the environmental conditions e.g. humidity and temperature, CO2 concentration etc. and the concrete's physiochemical conditions e.g. compressive strength, additives and surfacing etc., we further broke down the utilisation of concrete based on these specifications (see https://doi.org/10.5281/zenodo.4064803 (Wang et al., 2020)). The regional-specific calculations were then realised by regrouping the data based on their region-specific sources. Consequently, the regional and global uptakes can be calculated by aggregating each compressive strength class ( i ) 5 as 155 where the common symbols keep their meanings as defined previously and i c stands for the cement content of concrete. In short, on top of the regional cement production and/or clinker ratio data, other statistics necessary to carry out such regional calculations include (all regional) the proportion of cement used for making concrete (as opposed to mortar), the cement contents, the CaO content of clinker, the distribution of compressive strength class and the average thickness of different 160 concrete utilisations. Crucially though, diffusion coefficients of CO2 in concretes of the above specifications and the After their service life, concretes are usually demolished for either landfill or being reused. Reusing concrete at the end of its service life has been encouraged and envisaged to reduce the total emissions and increase the sustainability of the cement industry (IEA and WBCSD, 2018). However, the reusing rate of demolished concrete had been found to be very low at about 25% worldwide (Kikuchi and Kuroda, 2011;Yang et al., 2014). Demolition entails crushing of the bulk concrete structures so that the embedded steel structures can be easily extracted and recycled hence the end-product is usually in broken 170 piece. Therefore, the surface area exposed to the air dramatically increase during the demolition stage. As pointed out earlier (Eq. (3)), the exposed surface area is one of the key parameters that is positively correlated with the rate of carbonation, it is therefore expected that the carbon sequestered per unit time would increase with an increasing exposure. Again, we simplified the geometrical aspects of the calculations by assuming the demolished and crushed concrete parts ended up in spherical shapes so that the carbonation starts from the outer surface moving inwards radially (see Figure 2). Similarly, we considered the same 175 diffusion model to be applied for the carbonation process. Based on the survey of typical crushed cement particle sizes, we divided the distributions into three distinct groups according to their respective minimum (a) and maximum diameters (b) in the range, with respect to the maximum diameter ( i D0 ) that a particle will undergo full carbonation in compressive strength . The corresponding methods for calculating their carbonated fraction ( di F ) then are as follows: 180 6 The rationale is that ROW is mainly comprised of developing nations, hence it is more likely that the utilisation of concrete adopts similar patterns to China.
where di k and d t 7 are the diffusion coefficient of 'exposed to air' condition for compressive strength class I and the time 185 between service life and subsequent dealings. Besides, based on the survey data from literature for the particle size, we assumed a uniform distribution between a and b for each reginal subcategory.
Since carbonation during the demolition stage took place only in the bulk of concrete material where it remains uncarbonated after used in service, the fraction of carbonated concrete before demolition should be excluded from the calculation to avoid double counting. We assigned the total mass of consumed cement as mci 8 and the carbonated cement in 190 service life as mli (mli = di*Ai*ci as in Eq. (3)). Therefore, the total amount of CO2 uptake during the demolition stage ( can be calculated as: Carbonation during the secondary use stage that follows would be slower because a carbonate layer has formed at the particle surface previously. It might be less confusing to the readers to think of the demolition and secondary use stages as a whole with the diffusion process slowing down during the latter. Additionally, because of the high rates of landfill postdemolition, the diffusion processes are further retarded in the buried conditions 9 (Papadakis et al., 1991;Yoon et al., 2007).

200
Therefore, we introduce a lag time t  for it would take longer for the carbonation to reach the same depth ( di d ) when concrete 7 The average value was estimated to be 0.4 years worldwide (Pade and Guimaraes, 2007). 8 The cement consumed was taken the same as the cement produced. The discrepancies were sought after for certain years and considered in the uncertainty analysis. 9 particles are in the secondary use conditions compared with the demolition conditions ( di t ): and we have: 205 where the common symbols shared with Eq. (4) have the same meanings and si k stands for the diffusion coefficient during secondary use for compressive strength class i. By now, we can represent the combined carbonation depth of demolition and secondary use stages with all known variables as: where si t is the average time of the secondary use stage 10 and ti D is the maximum diameter that a particle will undergo full 210 carbonation in compressive strength class i in the demolition and secondary use stages combined. Similar to how we determined the fraction of carbonation previously, the fraction of further carbonation during the secondary use stage can be calculated by integration according to the same set of particle size criteria: Like for the demolition stage where double-counting was avoided by excluding the carbonated concrete during service, calculating the carbonation during secondary use should be based on the uncarbonated fraction of concrete after the service and demolition stages. Accordingly, the carbon uptake during the last piece of the concrete life cycle can be expressed as 220 follows: where di m stands for the mass of concrete carbonated during the demolition stage. Overall, Eq. (1) can be applied to obtain 10 Since the life cycle of concrete is assessed on a 100-year basis,  Additionally, since mortar carbonisation has not been quantified before as far as we are aware, we conducted experiments to measure the mortar carbonation rate coefficients and the proportion of CaO converted to CaCO3 of typical mortar cements 235 produced in China and used these measured datasets as being representative of the other regions owing to a lack of data. Like concrete, mortar carbonation processes were also simplified to a two-dimensional diffusion 'slab' model in which carbonation starts at the exterior of the slab and gradually moves inwards, and similarly, Fick's 2 nd Law was applied to determine the carbonation depth in the general form. However, mortar cement diffusion rates ( m K ) were shown to be higher than concrete which has a lower cement content, higher water/cement ratios, and finer aggregate grains (El-Turki et al., 2009). The total 240 mortar carbonation can be determined based on Eq. (2) and (11) with the corresponding proportion of CaO conversion ( 1  , see Eq. (14)) adjusted to the mortar situation as measured. Again, we assumed the diffusion front is equivalent to the carbonation front and the area behind the front was regarded fully carbonated.
The large exposure area and thin layers of mortar cement translate into rapid carbonation. We calculate annual mortar cement carbon uptake based on the proportion of annual carbonation depths of the utilisation thicknesses. The annual 245 carbonation of mortar used for rendering, plastering, and decorating is calculated as follows:   of CaO within fully carbonated mortar cement that converts to CaCO3. After the carbonation depths in adjacent years were determined, the annual carbonation percentage was obtained by the difference between adjacent years to the total utilisation thickness. Combined with the cement for mortar and the percentage of mortar for rendering survey data, the annual carbonation of rendering mortar is then quantified. Calculation for carbon uptake of repairing and maintaining cement mortar is similar to rendering, plastering, and decorating mortar, with differences lie in the utilisation thickness and the percentage of mortar for 260 repairing and maintaining.
In comparison to mortars for rendering and repairing, masonry mortar take a longer time to complete carbonation due to the partially exposed condition, thicker utilisation layers and their covering by rendering mortar on masonry wall surfaces.
Here, we classify masonry walls into walls with both sides rendered ( mbt C ), walls with one side rendered ( mot C ), and walls without rendering ( mnt C ). We conducted an extensive survey to collect data on the extents to which mortar rendering has been 265 applied to masonry walls in China and used the representative data of China for other regions due to a lack of data. The carbon uptake of masonry mortar can be calculated as an aggregate of the three subcomponents:  Based on the models outlined above, the calculation of masonry mortar carbonation is similar to rendering and 280 repairing mortar in that determining the annual carbonation is according to the proportion of carbonation depth. The carbonation of masonry mortar for walls with both sides rendered is as follows:   r is the percentage of masonry mortar with both sides rendered of total masonry mortar. Masonry mortar for walls with both sides first start carbonation from both sides of exterior rendering and then gradually moves inward. When the utilisation time of masonry mortar is shorter than the time required for full carbonation of rendering mortar on the masonry wall, there is no chance for 295 the underlying masonry to be exposed to CO2, hence carbonation should not happen; otherwise, the fraction of carbonated rendering mortar on the surface should be excluded from the calculation to avoid double-counting. If the utilisation time of masonry mortar is longer than the time required for full carbonation of rendering mortar but shorter than the building service life, we used the carbonation depth difference between adjacent years to the total thickness to show the carbonation fraction of masonry mortar with both sides rendered in year t. If the utilisation time of masonry mortar is longer than the building 300 service life, it was assumed that the left uncarbonized masonry mortar will be fully carbonated in one year due to the large exposure area post-demolition. Therefore, the fraction of masonry mortar carbonation after service life can be quantified by the difference between the fraction of masonry mortar carbonation in service life (that is, the carbonation depth during the service life to the total thickness) and the total masonry mortar carbonation of 100%. The calculation for the carbonation of masonry mortar for walls with one side rendered differ only in the carbonation depth calculation i.e. without rendering on one 305 side, CO2 directly contacts the bare masonry mortar, so that only the fraction of carbonated rendering mortar on one side was excluded. Similarly, for walls without rendering at all, CO2 directly contacts the bare masonry mortar from both sides so that the total carbonation depth is twice of the carbonation depth on one side.

Uptake assessments of construction wastes
Cement wastes mainly arise during construction and accounts for 1% to 3% of total cement consumption according to 310 construction budget standards (Zhou, 2003) and survey data (Lu et al., 2011). Most of this waste is in small pieces and will be recycled as back fill or landfilled after the completion of building projects, of these about 45% is concrete and 55% is mortar (Bossink and Brouwers, 1996;Huang et al., 2013). Here, we adhere to the breakdown of construction wastes into 2 components where wastecon C and wastemor C are the uptake of the corresponding component, respectively. Given the small piece sizes and hence large exposure area of the construction wastes, we made a few simplifications according to the literature survey: 1.
Assuming waste mortar completely carbonate in the first year; 2. Assuming waste concrete completely carbonate over the following 5 years (ranging from 1 to 10 years) (Bossink and Brouwers, 1996;Huang et al., 2013). Consequently, the carbon 320 uptake of construction wastes can be quantified by the annual carbonation fraction in line with the ratio of carbonation depths to the full carbonation depths. The expression of construction wastes carbonation as following: where ci W is the cement used for concrete in strength class i; con f is the loss rate of cement for concrete in construction stage, 325 cont r is the annual carbonation fraction of construction waste concrete, mi W is the cement used for mortar in strength class i, mor f is the loss rate of cement for mortar, mor r is the annual carbonation fraction of construction waste mortar. In short, in addition to the regional cement production, clinker ratio data, other statistics needed to conduct the calculation include the distribution of compressive strength class, the loss rate of cement for concrete and mortar in construction stage as well as the carbonation time of construction wastes. Crucially though, the latter two statistics, for which we collected the data on a regional basis, will dictate the amount of carbon uptake.

Uptake assessments of cement kiln dust
Cement kiln dust (CKD) is the major by-product of cement manufacturing process and has traditionally been considered as an industrial waste. Most of CKD is diverted to landfills and a small part is beneficially reused (Khanna, 2009;USEPA, 1993). CKD is composed of fine, powdery solids and highly alkaline particulate material, and is similar in appearance 335 to Portland cement (Seo et al., 2019). Given the very small particle size (predominantly ranges from a few microns to 50 μm and some coarse particles between 50-100 μm, (Kaliyavaradhan et al., 2020)), CKD full carbonation in landfill conditions can be achieved very rapidly within one year and indeed substantial carbonation even occurs within the first 2 days of reaction (Huntzinger et al., 2009a(Huntzinger et al., , 2009bSiriwardena and Peethamparan, 2015). Therefore, the carbon uptake by CKD is calculated as follows: where cem W is the cement production, CKD r is the CKD generation rate based on clinker production,

Yearly and cumulative uptake calculations
While the sectoral carbon uptake can be analytically estimated by the corresponding sectoral equations i.e. for concrete, mortar, construction waste and CKD, respectively, using aggregated regional datasets as the inputs, the regional carbon uptake was determined by aggregating all sectoral contributions but with disaggregated regional production/consumption and 350 diffusion/carbonation coefficient, concrete structure thickness, concrete strength distribution, mortar utilisation distribution, waste particle distribution and CKD generating rate data, among others, as the model inputs. Consequently, the world total uptake can be divided up according to the usage of the cement produced as well as where the cement was produced/consumed 11 .
For mortar cement, we explicitly showed how to determine the annual carbonation from Eq. (12) to (14) and Eq. (16) to (18). Basically, for the carbon uptake of a specific year t, we can apply a simple subtraction of the cumulative values between 355 adjacent years as: so that each year's contribution to the total carbonation can also quantified. This way, we will be able to visualise the time lag in the carbonation process in that the uptake of a specific year t is not limited to the cement produced in the same year. 360

Uncertainty analysis
Based on the kinetic models described in previous sections, the annual regional carbon uptake was calculated by aggregating the contributions from individual types of cement. Likewise, the annual global carbon uptake was obtained from regional aggregation. It should be noted, though, that a Monte Carlo analysis method with 26 variables (see https://doi.org/10.5281/zenodo.4064803 (Wang et al., 2020)) was applied to evaluate the carbon uptake at each level, hence 365 the annual median at a higher level (i.e. regional wrt. cement type and global wrt. regional) is not equal to the sum of its sublevel components. The variables associated with the estimates are mostly in common with our previous study (Xi et al., 2016), with the only difference being the distribution of the clinker ratio. Previously, the clinker ratio was set to range from 75% to 97%, in a Weibull distribution with shape and scale parameters of 91.0% and 25 for the years of 1990-2019. In this research, for China, based on previous studies and local survey data, we adjusted the corresponding uncertainty range for the 370 1990-2019 period. Specifically, for 1990-2004, the range of coefficient value of clinker ratio was set to 10%-20%. In this range, the pseudo-random numbers were generated with a uniform distribution then multiplied by the mean values of clinker ratio to obtain the corresponding standard deviation. As such, the normal distributed random clinker ratio values were created.
For 2004-2019, the random errors were calculated within the range of ±5% of the mean values with a uniform distribution.
For 1930 -1989, the clinker ratio distribution was unchanged.

375
On the other hand, emission estimates are subject to uncertainties due to incomplete knowledge of activity levels and emission factors. In order to assess the uncertainties in our results more thoroughly, we conducted a comprehensive analysis of regional emission estimates. Following the method of previous studies (Tong et al., 2018;Zhao et al., 2011), we performed a Monte Carlo analysis that varied key parameters including cement production, clinker ratio, and emission factors. The term "uncertainty" in this study refers to the lower and upper bounds of a 95 % confidence interval (CI) around our central estimate 380 i.e. median. All of the input parameters of activity levels and emission factors, with corresponding statistical distributions, were fed into a Monte Carlo framework and 10,000 simulations were performed to analyse the uncertainties of estimated CO2 emissions. For the uncertainties of the regional process CO2 emission estimates, national average emission factors were derived from previous studies and local survey databases (Andrew, 2018;Hanle et al., 2006;Shen et al., 2016) and we assumed these activity rates are normally distributed, with coefficients of variations (CV i.e. the standard deviation divided by the mean) 385 ranging from 0.05 to 0.2 based on the specific data sources and year. Furthermore, the ranges of parameter values also vary by country in part due to the quality of their statistical infrastructure.

Aggregated regional and global process emission
With the continued increase in the production of cement and associated clinker globally in the past few decades, the 390 process CO2 emissions correspondingly have been increasing with limited abating measures (i.e. carbon capture and storage, CCS). According to our estimates, by 2019, the global process CO2 emissions reached 1.57 Gt yr -1 (1.42 Gt-1.86 Gt, 95% CI) (see Figure 4a), equivalent to about 25% of the total CO2 emissions from industrial activities in 2018 (Tong et al., 2019). emissions correspondingly increased by about 49 times, which is actually slightly slower than the increase in production. This is partly due to the relative decreases in average clinker ratios (from ~89% in 1930 to ~70% in 2019). Meanwhile, the regional attribution of such an increase changed significantly during the same period. As we can see in Figure  As mentioned in 2.2, there are other studies estimating the process emission based on high-resolution, national-level clinker ratio data. Andrew (2018) reported the process emission in 2017 to be 1.48 ± 0.20 Gt CO2 and that of aggregated 1928-410 2017 to be 36.9 ± 2.3 Gt CO2. Using our simpler regional based approach gives the 2017 process emission to be 1.51 Gt CO2 . This is clearly illustrated in Figure 5a where the area representing each region denotes the amount of uptake. In the US and Europe, since the cement stock per capita has 420 reached saturation (Cao et al., 2017) and the concrete structures generally have long service lives (70 and 65 yrs for Europe and the US on average, respectively) relative to the life cycle (i.e. 100 years) considered in our model, it is conceivable that the absolute uptake in these two regions only have been increasing mildly after 1980s, which is in drastic contrast to the 'exponential' rise observed for China (see Figure 6). In terms of ROW, the increase in uptake has been somewhat intermediate  (Wang et al., 2020)) and literature (Lutz and Bayer, 2010;Winter and Plank, 2007).

Characteristics of cement carbon uptake
One of the limitations of natural carbonation for carbon capture is that it is a slow process hence speeding up the chemical

22
processes involved is the key to realise tangible impacts on mitigating CO2 emissions. This is also the case for cement materials especially concrete structures, which took up the majority of their utilisation. Therefore, the carbon uptake by concretes, before demolition, has been persisting during their lifetimes. This is evident in Figure 7 where the cements materials (mainly concretes) 445 consumed in a given decade (colour-coded) still made contribution to carbon uptake decades later. In spite of this feature, more than 75% of the total uptake was attributed to, based on our estimates, the cement materials produced/consumed after the 1990s. This is in line with the trend of process emission growth i.e. 75% vs 73% in the same 1930-2019 period. The difference 12 can be accounted for by the dynamic processes and the varying durations of the stages involved in the life cycles, as considered and implemented in the uptake models. This contrasts with the immediate process emission process. It is also suggested in 450 Figure 7 that a surge in uptake occurs at the demolition stage because of the significant increase in fresh surface area.  12 It is not necessarily the case that the fraction of uptake is smaller than that of the process emission for post-1990 period. Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License.

Figure 8
The annual carbon uptake (median) of concrete produced/consumed in Europe and Central Eurasia, attributed to the current: the concrete produced/consumed in that year and to historical: the concrete produced/consumed in the years prior to that year.

465
All the original datasets used for estimating the emission and uptake in this study and the resulting datasets themselves from the simulation, as well as the associated uncertainties are made available by Zenodo at https://doi.org/10.5281/zenodo.4064803 (Wang et al., 2020).

Conclusions
Estimating CO2 uptake of cements is essential for evaluating the real environmental impact of the cement industry.

470
Previous efforts were limited by data availabilities and incomplete accounting for other cement materials other than concrete.
From a historical perspective, while mortar had absorbed more CO2 than any other types of cement, more uptake had occurred https://doi.org/10.5194/essd-2020-275 in China than in any other country, owing to its dominant cement production/consumption position in recent decades (>43 % from 2000 to 2019). The microscopic kinetic processes dictate that CO2 uptake of cement is a dynamic process such that legacy absorption from cements produced in the past should not be omitted. Overall, post-1990 era sees more than 75% of the total uptake estimated. As a revision to our previous work (Xi et al, 2016) where the clinker ratios were likely to have been overestimated, a dynamic clinker ratio approach was adopted to reflect the recent technological changes in the industry despite limited to China and India only. Besides, the dynamic clinker ratios were also applied in re-evaluating the process emissions.
The compounded results suggest that the cumulative CO2 offset reached approx. 52% as of 2019 (see Figure 4a).
This dataset and the estimate methodology can serve as a set of tools to assess the emission and, more importantly, the 480 uptake of CO2 by cement materials during their life cycles. Given cement demand is projected to remain increasing to satisfy society developments globally, future work is needed to increase the accuracy of the uptake estimates, crucially, by utilising the direct clinker production data where possible and obtaining spatially resolved conversion factors determined by experiments.