The soil samples were collected from 148 sample sites in 41 cities that were evenly distributed across mainland China except for the Tibetan Plateau during 2013–2014 (Fig. 1). Depending on the city size, multiple sample sites were identified in each city. Each site belonged to a separate city district, i.e. the soil samples were taken from 148 different city districts across China. The sample sites included various ISA types (roads, elevated piers, buildings, etc.) and PSA types (trees, shrubs, herbs, bare ground, vegetable plots, etc.). Detailed descriptions of the cities and sample sties can be found in Ding et al. (2023).

At each sample site, three representative ISA sampling plots, more than 10 m apart from each other, were randomly selected. In addition, three paired sampling plots in adjacent PSAs were randomly selected for comparison. In each plot, a 100 cm depth profile pit was dug, and the soil profile was sampled at 20 cm intervals to the 100 cm depth with a 100 cm3 ring knife. Our study across China found that most of the Ekranic (sealed) Technosol profiles have a clear boundary between the building material layer and the soil. Where the boundary is unclear, we treated the topsoil with a high amount of hard building materials, where artefacts >0.15 mm accounted for over half of the soil volume, as the building material layer. We only took samples in the soil below the building material layer. Samples with notable additions of anthropogenic artefacts, e.g. coal fly ash, mixed in the soil were discarded. Following the protocol of China's National Soil Surveys, the visible non-soil artefacts in the remaining soil samples, such as fragmentations of bricks, glasses, stones, roots, and so forth, were picked out and discarded (Shi et al., 2004). A total of 4356 soil samples were eventually collected from 888 soil profiles. These samples (ID# XJBIZC0001–XJBIZC4356) are currently stored in the herbarium of the Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences.

To facilitate spatial analysis, we divided the country into six subregions: the northeast, north, northwest, east, south, and southwest, according to geography, climate, and socioeconomics following Ding et al. (2022). To estimate the NISA storage in each subregion, we multiplied the mean NISA density in the region by the region's ISA land area, which was derived from the 30 m resolution ISA map of mainland China (Zhang et al., 2020). Then, the NISA stock of all subregions were added up to estimate the national NISA storage.

In this study, soil bulk density (BD) and N content were measured for each soil sample. Soil samples inside the ring knife were dried at 105 ∘C, and soil bulk weight (BD) (g cm-3) was measured, while the rest of the samples were air dried and passed through a 0.15 mm sieve, and the N content (g kg-1) was measured by Kjeldahl digestion (Bremner and Mulvaney, 1982). The N density (kg m-2) per 20 cm soil layer was calculated according to Eq. (1), and the N density for the entire 100 cm profile was obtained by summing the N density per 20 cm soil layer (Eq. 2): 1Ni=NCi×BDi×20100,2N100cm=∑i=1nNi, where N represents N density (kg m-2), i∈[1,5] represents soil layer (each 20 cm in thickness), NC is N content (g kg-1), and BD is soil bulk density (g cm-3).

A paired t test (two tailed) was used to determine the difference between NISA and NPSA and the difference between C:NISA and C:NPSA (C:N ratio of the PSA). The C:N stoichiometry, i.e. the C:N ratio, shows the connection between the C process and N process. An extremely high C:NISA in comparison with the reference C:NPSA indicates C–N decoupling due to soil sealing (Raciti et al., 2012).

The SOC density of the samples was reported in a previous study (Ding et al., 2022). We noticed that some research (Hu et al., 2018; Pereira et al., 2021; O'Riordan et al., 2021) used the ratio between total C and total N to investigate the C:N stoichiometry in ISA soil. However, the content of soil inorganic C under impervious surfaces is likely altered by anthropogenic C from construction materials, and black C (the soot or carbonaceous products formed during the incomplete combustion of biomass and fossil fuels) (He and Zhang, 2009; Zhao et al., 2017; Zhu et al., 2019; O'Riordan et al., 2021). In this study, we used the ratio between SOC and N to investigate the soil C:N stoichiometry, just like most soil studies in both ISA (Wei et al., 2014a; Raciti et al., 2012; Piotrowska-Długosz and Charzyński, 2015) and PSA (Lu et al., 2023; Schroeder et al., 2022; Yang et al., 2021).

We further investigate whether soil sealing may influence the variations of C:N stoichiometry at the national scale. According to the urban ecosystem convergence theory, intensive human disturbances (e.g. soil sealing) could reduce variations in soil property at large scale (i.e. among different cities) even if the intensively disturbed areas may have similar or higher variations in soil properties at city scale compared with the less disturbed areas (e.g. PSA) (Pouyat et al., 2003). To evaluate this theory, we compared the mean inter-city C:N stoichiometry dissimilarity and the mean intra-city C:N stoichiometry dissimilarity between the ISA and PSA. The inter-city dissimilarity (or regional scale dissimilarity) measured the Euclidean distance (Eq. 3) in C:N between each pair of different cities, while the intra-city dissimilarity (or local scale dissimilarity) measured the Euclidean distance in C:N between each pair of sampling sites within the same city, all combinations included. If the urban ecosystem convergence theory was correct, we expect to see ISA having lower inter-city C:N dissimilarity than PSA, but higher or similar intra-city C:N dissimilarity than or to PSA: 3Euclideandistance=C:Ni-C:Nj2, where C:Ni and C:Nj are the soil C:N ratios of site i and site j, respectively, when measuring the intra-city dissimilarity, or the city-averaged C:N ratios of city i and city j, respectively, when measuring the inter-city dissimilarity.

Unlike other studies that focused on topsoil, our multiple–layer soil sampling data made it possible to study the vertical pattern of NISA to a 100 cm depth. The proportions of N stored in the 0–20 cm depth, 0–40 cm depth, 0–60 cm depth, and 0–80 cm depth to the total (100 cm depth) N stock in each sample profile were calculated and plotted against the soil depth to reveal the vertical distribution pattern of NISA and NPSA. Based on these data, we could model how N storage changed with soil depth. According to Yang et al. (2007), 46 % of the N stock (in 1 m depth) of natural soil is stored in the top 0–30 cm soil, and 68 % of the N stock is stored in the top 0–50 cm, translating into a power function fitting model: 4NNatural%d=-0.0074d2+1.7378d=1.7378-0.0074d×d, where NNatural%d is the proportion of total N stock (in 100 cm depth) stored to depth d cm in natural soil in China. The equation shows that the NNatural%d does not increase linearly with soil depth; instead, its increasing rate (i.e. 1.7378–0.0074 d) reduces with soil depth d. This pattern indicates the natural soil N does not have homogeneous N density throughout the soil profile; it decreases with depth.

Our large-scale soil survey made it possible, for the first time, to investigate the correlations between soil N and various environmental factors so as to identify the climatic, ecological, geographical, and socioeconomic factors that may control or influence the N and C:N dynamics in sealed soil. We selected 15 indicators to investigate the factors associated with NISA, including mean temperature, annual precipitation, background NPP (averaged natural ecosystem NPP in a 5 km buffer outside the city), C:NPSA and NPSA, longitude, latitude, elevation, population density, built-up area in a city, urbanization rate as indicated by the fraction of the built-up area that expanded after 2002, ISA coverage in built-up areas, greenspace coverage in built-up areas, per capita greenspace, city GDP, and per capita GDP. We also investigated the correlation between soil BD and the NISA content.

Gridded datasets of environmental factors, including mean annual temperature (Fig. 2a), annual precipitation (Fig. 2b), and elevation (Fig. 2d) at 1 km resolution, were obtained from the Data Center for Resource and Environmental Sciences, Chinese Academy of Sciences (http://www.resdc.cn/, last access: 20 July 2021). The national NPP (1985–2015) estimates at 1 km resolution was obtained from the Digital Journal of Global Change Data Repository (https://www.geodoi.ac.cn/, last access: 20 July 2021) (Fig. 2c). Statistical datasets include the Ministry of Housing and Urban-Rural Development of China (http://www.mohurd.gov.cn/, last access: 23 April 2021) urban built-up area, population density, built-up area green space rate, per capita built-up area green space, the National Bureau of Statistics of China (http://data.stats.gov.cn/, last access: 23 April 2021) total urban GDP, and per capita GDP. We used correlation analysis to investigate the relationships. If the variables were normally distributed and linearly correlated, then the Pearson's correlation (two tailed) was applied. Otherwise, Spearman's correlation (two tailed) was used.

The national mean NISA density in the 100 cm soil profile was 0.59±0.35 kg m-2 (mean ±1 SD), ranging from 0.08–1.88 kg m-2 with a median value of 0.48 kg m-2. Paired t tests showed that the NISA was significantly (approximately 30 %) lower (P<0.01) than the reference NPSA (Fig. 3a). Moreover, the NISA was lower than the reference NPSA at all soil depths and in all subregions of China (Fig. 4a). The national total NISA stock was about 98.74±59.13 Tg N.

C:NISA (10.33±2.62) was significantly lower than C:NPSA (10.93±3.19) (Fig. 3b). Moreover, NPSA and SOCPSA were significantly correlated (R=0.893, P<0.01), and NISA and SOCISA were also significantly correlated (R=0.926, P<0.01). There were no signs of C–N decoupling according to our data.

To facilitate spatial analysis, we divided the country into six subregions: northeast, north, northwest, east, south, and southwest, according to geography, climate, and socioeconomics (Fig. 1a) (Ding et al., 2022). The highest NISA density of 0.84±0.45 kg m-2 was found in the east, while the lowest NISA density of 0.51±0.20 kg m-2 was found in the south (Fig. 3c). Notably, the urban soil N densities, both in ISA and PSA, in the east were significantly higher than that of the other subregions, except for the southwest where the variations of soil N densities were extremely high (Fig. 3c). As a result, the east subregion accounted for the largest share (34 %) of the NISA stock in China (Fig. 3d).

Figure 4 shows the ISA soil samples had lower inter-city C:N dissimilarity (1.86±1.40 vs. 2.43±2.15) than PSA but similar intra-city C:N dissimilarity to PSA. This pattern indicates that although the ISA soil and the PSA soil had similar variations in C:N stoichiometry at the local scale (within a city), the C:N variations at national scale (among the cities) were reduced for the ISA soil, possibly due to the intensive human disturbances on ISA soil as predicted by the urban ecosystem convergence theory (Pouyat et al., 2003).

The spatial trend analysis of the NISA showed a slow decline followed by a rapid increase in the north–south direction and a rapid decline followed by a rapid increase in the east–west direction (Fig. 5a). The spatial trend analyses of the NPSA produced similar concave lines in both the north–south and the east–west directions but with a more drastic initial decline in the north–south direction and a flatter trend in the east–west direction (Fig. 5b). The spatial trend analysis of C:NISA showed a rapid increase in the north–south direction and a rapid decrease followed by a slow increase in the east–west direction (Fig. 5c). The spatial trend analysis of C:NPSA showed a slow increase in the north–south direction and a slow decrease in the east–west direction (Fig. 5d). According to the spatial trend analyses, the change rate of C:NISA was significantly higher than that of C:NPSA, which was consistent with the results of inter-city C:N ratio dissimilarity analysis (Fig. 4).

In this study, the vertical profiles of soils under ISAs were systematically sampled and analysed at 20 cm intervals to a 100 cm depth, and the storage of NISA increased linearly with soil depth (Fig. 6). This linear distribution pattern was evident at the national scale (R2=1, P<0.001), and the vertical distribution pattern of NISA in China can be described by a linear model (Eq. 5), 5NISA%d=1.0324d, where NISA %d is the percentage of the N stock (to 100 cm depth in total) located in the top d (cm) depth of the soil.

Latitude, temperature, NPP, and C:NISA were non-normally distributed, so we analysed them using Spearman's correlation (two tailed), and the correlations of the remaining variables were analysed using Pearson's correlation analysis (two tailed). The impacts of climate and geographic factors were confirmed by correlation analyses, which showed NISA to be negatively correlated with temperature (R=-0.486) (Table 2). In addition, the NISA had a positive correlation with the N of urban PSA (R=0.715) and a negative correlation (R=-0.34) with the urbanization rate as indicated by the fraction of the newly expanded ISA since 2002 (i.e. fnew_ISA=ISA2015-ISA2002ISA2015). Surprisingly, we did not find significant correlations between NISA and common environmental drivers like precipitation and NPP at the 95 % significance level, although the NISA had a weak negative correlation with precipitation (R=-0.268) at the 90 % significance level.

The C:NISA was negatively correlated with both precipitation (R=-0.620) and temperature (R=-0.561) but positively correlated with latitude (R=0.513) (Table 2). In addition, the C:NISA had a positive correlation with the N of urban PSA (R=0.515) and a negative correlation (R=-0.516) with the NPP.

Our results were comparable to, or moderately lower than, the previously reported topsoil (0–20 cm) NISA contents and densities in Chinese cities (Table 1), including Beijing (0.34±0.06 g kg-1 vs. 0.37–0.61 g kg-1) (Hu et al., 2018; Zhao et al., 2012), Nanjing (0.13±0.15 g kg-1 vs. 0.49 g kg-1) (Wei et al., 2014b), and Yixing (0.15±0.01 kg m-2 vs. 0.25 kg m-2) (Wei et al., 2014a). Our observed NISA content (0.4 g kg-1) in the 20–40 cm soil layer in Beijing was also comparable to the reports by Zhao et al. (2012) (0.26–0.42 g kg-1). Outside China, the reported topsoil (0–10 cm) NISA density in Greater Manchester, UK (0.081 kg m-2), was comparable to our estimated mean NISA density in 0–10 cm (0.07±0.04 kg m-2) in China (O'Riordan et al., 2021), but the reported NISA in New York, USA, and Toruń, Poland, was much lower than our results and the reports from other Chinese city studies (Table 1) (Raciti et al., 2012; Piotrowska-Długosz and Charzyński, 2015).

Compared with previous assessments of China's soil N stock, NISA accounted for 0.96 %–1.47 % of the total soil N pool in China (Zhang et al., 2021; Xu et al., 2020). The NISA pool size (98.74±59.13 Tg) exceeded the vegetation N of scrubland (8.1–50 Tg) (Xu et al., 2020) and grassland (48.8 Tg) (Zhang et al., 2021) in China. The NISA density (0.58±0.12 kg m-2) was lower than that of natural soil and equivalent to 53 %–69 % of the national average (Yang et al., 2007; Xu et al., 2020). The N densities in ISA soil (0.59±0.35 kg m-2) were lower than those in other ecosystems, such as forest (1.29 kg m-2), farmland (1.13 kg m-2), and grassland (1.11 kg m-2) (Xu et al., 2020), indicating that ISA construction resulted in N loss. Previous studies ignored the impacts of ISAs (Tian et al., 2006; Yang et al., 2007) and thus might have overestimated China's soil N pool size.

Our estimated C:NISA (10.33±2.62) in China matched the previously observed C:NISA (10.8) in Yixing, China (Wei et al., 2014a). It has been suggested that different terrestrial ecosystems may have similar C:N ratios (Yang et al., 2021). This study showed that the C:NISA (10.33±2.62) was only slightly lower than the C:NPSA (10.93±3.19) in urban ecosystems but much higher than the C:N ratios of natural ecosystem soils such as forests (8.21), croplands (8.18), and grasslands (7.7) (Xu et al., 2020; Tang et al., 2018). Therefore, it is possible that the C:N stoichiometry could remain relatively stable within the same land-use type (natural ecosystems, urban areas, etc.) but might differ significantly among different natural or human-disturbed land-use types.

The above comparison indicates that ISA soil has a higher C:N than natural soils. A study in New York City reported that the N density in the ISA was 95 % lower than that in the PSA, leading to an extremely high soil total C : total N ratio (Raciti et al., 2012; Majidzadeh et al., 2017). Therefore, Raciti et al. (2012) suggested that paving decouples the C and N cycles. Our observations, however, showed that the soil N of ISA was only 30 % lower than that of PSA, and the C:NISA was lower than the C:NPSA in China. Furthermore, there was a significant positive correlation (R=0.926, P<0.01) between N and SOC in ISA soil. Similarly, Wei et al. (2014a) found that C:NISA was lower than C:NPSA in Yixing, China, and O'Riordan et al. (2021) found a significant positive correlation between N and C in ISA soil in Greater Manchester, UK, even though they also observed an increased total C : total N ratio in ISA soil compared with PSA soil. There were no signs of C–N decoupling according to our data and others (O'Riordan et al., 2021; Wei et al., 2014a). It is possible that the extremely high C:N ratio observed in previous studies might be merely caused by anthropogenic C inputs that partially compensated for SOC loss during land conversion (O'Riordan et al., 2021). Because the construction materials could add large amounts of inorganic C into soil, it is preferable to investigate the C:N stoichiometry under ISAs with the C:N ratio rather than the total C : total N ratio. This study highlights the important role of N in urban biogeochemical research, which helps to prevent us from being confused or misled by the complex C dynamics in urban soil due to anthropogenic C inputs.

The spatial distribution pattern of soil N was significantly correlated with climate factors such as temperature and precipitation in natural ecosystems (Yang et al., 2007). In general, the soil N in China's temperate and subtropical ecosystems were negatively correlated with temperature (Lu et al., 2017). Similarly, our study found a negative correlation between NISA and temperature. There was no significant correlation between precipitation and the soil N in natural ecosystem, except for dryland where a positive correlation has been found (Lu et al., 2017). We did not find significant correlation between precipitation and NISA at the 95 % confidence level, although there was a weak negative correlation at the 90 % confidence level. Previous study showed the SOCISA was also negatively correlated with precipitation, and it was suggested that the observed soil biogeochemistry (SOC, nutrient content, etc.) under impervious surfaces was mainly determined by the losses (especially in topsoil) during land conversion (Majidzadeh et al., 2018; Cambou et al., 2018; Edmondson et al., 2012). Higher precipitation leads to higher soil nutrient loss during land conversion (Ding et al., 2022). The relatively weak correlation between NISA and precipitation (compared with the correlation between SOCISA and precipitation) as well as the negative correlation between C:NISA and precipitation might indicate that the N loss during land conversion was not as significant as the loss in SOC. It is also possible that the high N deposition in urban ISA might somehow replenish the NISA pool.

NISA was not correlated with background NPP but positively correlated with the soil N in the adjacent urban PSA. This pattern agrees with the previous report that the SOCISA was mainly influenced by the SOC in the adjacent urban PSA rather than the background SOC and NPP (Ding et al., 2022). However, there was a negative correlation between C:NISA and background NPP.

The soil C:N ratio could be a more stable parameter (Yang et al., 2021). Tian et al. (2010) found the soil C:N ratio was relatively stable among climate zones in rural ecosystems in China. It has been observed that the soil stoichiometric characteristics in China are influenced by geographical parameters such as altitude and latitude (Sheng et al., 2022). Lu et al. (2023) found a lower C:N ratio at higher latitudes in China, suggesting a positive correlation between C:N and temperature in natural ecosystem soils. Our study, however, found that the C:NISA ratio increased with latitude and that there was a significant negative correlation between the C:NISA ratio and temperature. The soil C:N ratio of natural ecosystems is influenced by plant litter input and N uptake. Ecosystems in warmer regions have higher NPP, resulting in higher inputs of litter with a high C:N ratio (compared with the soil C:N ratio) and higher N uptake by roots, thus reducing soil inorganic N. Therefore, the C:N ratio is positively correlated with temperature in natural ecosystems. However, the C:N ratio under impervious surfaces is solely determined by the relative mineralization rate of C and N. It seems that soil ecosystems have a higher retention capacity for N than for C (C fixation is unlikely to be found in sealed soil). Therefore, while both the soil NISA pool and the SOCISA pool decrease when the temperature increases, the net N mineralization rate is lower than the C mineralization rate, leading to a negative correlation between the C:NISA ratio and temperature.

Our study and Tian et al.'s (2006) study on China's soil N had the same subregion zone design. However, we found that the urban soil (both the ISA and PSA) in the eastern zone had the highest N density while Tian et al. (2006) found that the rural soil N density in the eastern zone was among the lowest in the country. The relatively high precipitation and temperature in eastern China may lead to a high SOM decomposition rate and nutrient leaching rate, which explains its low rural soil N density (Tian et al., 2010). However, the eastern region also has been the most developed region in China for the past several centuries. Its cities have had a high population density and long urbanization history. The long-term intensive human activities might be leaving a profound footprint in the soil biogeochemical processes, significantly elevating its N content. This finding, together with the relatively low inter-city C:N variations and dissimilarities in the ISA (see Sect. 3.2), indicates that intensive human disturbances might override the natural environmental effects in shaping the regional distribution pattern of soil N processes, a perspective further confirmed by the urban ecosystem convergence theory (Pouyat et al., 2003).

The vertical N distribution patterns are also different between the ISA soil and rural soil. Unlike the vertical distribution pattern of natural soil N, which decreased with depth and should be modelled with a second-degree polynomial fitting function (Eq. 4), the vertical distribution of NISA was relatively homogeneous throughout the soil profile and could be modelled with a linear function (Eq. 5; Fig. 6). The SOC under impervious surfaces had a similar vertical distribution pattern (Ding et al., 2022). The unique vertical pattern reflects the effect of human disturbance during land conversion, which led to topsoil soil organic matter and nutrient loss, and reduced the SOC and N gradient throughout the soil profile (Majidzadeh et al., 2018; Cambou et al., 2018; Edmondson et al., 2012).

N plays an important and complex role in natural ecosystems. Soil microbial C use efficiency is negatively correlated with C:N (Schroeder et al., 2022). Incorporating N into earth system models can improve the accuracy of C cycle estimates (Fleischer et al., 2019), and a good description of N can help understand and predict the patterns and mechanisms of global C dynamics (Zhang et al., 2021) and provide a reliable basis for exploring how geochemical cycles are coupled.

For a long time, knowledge of biogeochemical cycles under impervious surfaces has been a major gap in urban biogeochemical research. Until recently, the size and pattern of NISA pools and their contributions to the global N cycle have remained unclear. Our research, which is the first national-scale study on NISA and C:NISA, helps to fill this gap by improving our understanding of the special pattern of soil N under impervious surfaces. Such information is necessary when assessing urbanization impacts on global C and N cycles (Lorenz and Lal, 2009).