Introduction

Cities currently host 56% of the global population, projected to rise to nearly 70% by 2050 (UN-Habitat 2022). The rapid pace of urbanization, alongside climate change impacts, underscores the imperative of ensuring urban sustainability, resilience, and livability. Urban Green Space (UGS) emerges as a promising solution to mitigate adverse urbanization effects such as urban heat islands, flooding, and pollution, thereby enhancing human well-being in urban areas (Endreny 2018; Ren et al. 2022; Wang et al. 2022a). However, issues of inequality in UGS distribution and accessibility persist across various scales, from local to global (Dong et al. 2022; Wu et al. 2023). Informed UGS planning is integral to urban human well-being, crucial for urban sustainability by 2030, aligning with the United Nations 2030 Agenda.

While creating UGS serves as a promising Nature-based Solution (NbS) to enhance human well-being in urban areas, its viability is subject to debate within the context of urban densification. Urbanization tends to deplete natural and productive landscapes, diminishing nature's contributions to people (Foley et al. 2005; He et al. 2014; Zhao et al. 2006). The Kuznets Curve posits a threshold level of urbanization where negative urbanization effects decline, necessitating careful urban planning to balance urban development with environmental and social factors (Dinda 2004; Yang et al. 2013). Urban densification, emphasizing reduced urban expansion and increased land multifunctionality, offers a viable strategy to limit adverse impacts on regional landscape sustainability (Cao et al. 2023; Haaland and van den Bosch 2015). However, this approach may allocate less land for UGS creation and encroach upon existing UGS for construction, limiting urbanites' exposure to UGS and its ecosystem services (Haaland and van den Bosch 2015; Tappert et al. 2018). Balancing UGS creation with urban densification introduces a trade-off in urban planning, despite both aiming to foster sustainable development. Given the escalating global trend of population migration to urban areas, meeting the growing residents’ demand for UGS within the urban densification framework is crucial for advancing landscape sustainability.

The primary driver behind urban densification is that nearly half of global urban expansion results in cropland loss, impacting grain production and food security (Huang et al. 2020; Kuang et al. 2022a). This phenomenon is particularly evident in recently urbanized countries like China, where two-thirds of urbanization directly led to cropland consumption between 2003 and 2016, contributing to a 2% decline in the national grain self-sufficiency rate from 1992 to 2015 (He et al. 2017; Huang et al. 2020; Qiu et al. 2020). Satellite observations reveal that nearly 38% of globally expanded urban areas consist of UGS from 2000 to 2020, reaching 40‒45% in North America and Europe (National Remote Sensing Center of China 2020). It is, therefore, deduced that amidst rapid urbanization, UGS expansion primarily promoted urban greening worldwide (Fig. 1). Global UGS coverage observed a rise from 28 to 33% over 20 years, providing a foundation to enhance global city resilience and livability (National Remote Sensing Center of China 2020). However, the impacts of UGS expansion on cropland remain unclear, potentially undermining landscape sustainability by creating a trade-off between efforts to enhance urban human well-being and foster agricultural ecosystem services (Wu 2021).

Fig. 1
figure 1

The proposed causal link between cropland and grain production loss resulting from urban green space expansion

Full size image

Scholars and decision-makers advocate for further UGS coverage enhancements in response to urban environment degradation, evident in urban greening schemes like NbS and the 3-30-300 strategy (Browning et al. 2024; Iungman et al. 2023). However, if future UGS creation primarily depends on the expansion of urban areas (i.e., UGS expansion approach), it may lead to additional cropland loss and reduced grain production. Global concerns are escalating due to agricultural loss, land degradation, hazards, and geopolitical conflicts, heightening the urgency of prioritizing cropland protection and food security (Cui and Shoemaker 2018; Lin et al. 2023). While UGS expansion is effective in enhancing urban greenery, there is a need for empirical evidence to support the establishment of a safety boundary for UGS expansion, aimed at minimizing its impacts on cropland protection and grain production.

Building upon previous studies (Huang et al. 2020; Kuang et al. 2021), we propose a hypothesis suggesting that the loss of grain production contributes to increases in UGS coverage. This contribution can take two possible forms: linear and non-linear, as depicted in Fig. 1 (Peng et al. 2017). Under ideal conditions, the contribution curve exhibits concavity, reaching a saturation point of grain production loss. Managing UGS coverage at this point can mitigate grain production loss and maximize its contribution, thereby informing landscape and urban planning strategies that promote landscape sustainability. Conversely, the linear contribution curve may show an unavoidable trade-off between grain production and urban greening. In such cases, future approaches to UGS expansion should be minimized or even abandoned, as addressing the impact on food security becomes challenging.

This study aims to identify a leverage point for UGS expansion that maximizes UGS exposure while minimizing its impact on grain production. Focusing on China, where the central government prioritizes strict control of urban expansion and safeguarding cropland, this study addresses three main issues: (1) Identifying patterns of cropland loss due to UGS expansion, 2000‒2020; (2) Assessing the impact of UGS expansion on grain production during the same period; and (3) Strategically mitigating this impact at the national level by 2040. The findings are integral for achieving integrated management of urban human well-being and surrounding landscapes through an informed UGS expansion strategy that aligns with the principles of urban densification.

Materials and methods

Study design

In China, urban expansion, along with its associated cropland and grain production loss, and urban greening (characterized by the growth of UGS area), has garnered global attention since the twenty-first century (Huang et al. 2020; Kuang et al. 2021; Tu et al. 2023a). The critical issues of cropland loss, soil degradation, and water limitations have elevated food security concerns at a governmental level (Cui and Shoemaker 2018). In response, the Chinese government implemented the “three zones delineated by three lines for land use” initiative in 2019, outlining red lines for the protection of permanent basic cropland, as well as control lines for urban development (Zhang et al. 2019). This initiative aims to reduce extensive urban expansion in China, with urban densification emerging as the primary direction for urban development. However, studies reveal the persistent inequality in UGS exposure amid rapid urbanization in China, prompting calls for creating UGS to enhance human well-being in urban areas (Dong et al. 2022; Song et al. 2021). In this context, an investigation into China's experience provides a valuable global reference point for understanding the trade-off between UGS expansion and grain production.

Our study focuses on the urban patch level across China for 2000‒2010 and 2010‒2020. Criteria for selecting urban patches (i.e., urban areas) include: (1) an area exceeding 6.25 ha (equivalent to one pixel of urban land cover data) and (2) continuous expansion during the investigation period. We analyzed 3402 urban areas, totaling 9.58 Mha in 2020, categorizing them into six geographical regions and five urban size groups (Fig. 2a). The six regions comprise Northeastern China (NEC), Northern China (NC), Northwestern China (NWC), Southwestern China (SWC), Central-southern China (CSC), and Eastern China (EC) (Jia et al. 2021). Urban size groups are classified by urban population according to government standards: I [< 0.5 million (M)], II (0.5‒1 M), III (1‒5 M), IV (5‒10 M), and V (> 10 M) (Dou and Kuang 2020).

Fig. 2
figure 2

The investigated urban areas and the proposed relationship between grain loss and UGS coverage. a The various categories of investigated urban areas. b Examples of cropland loss due to UGS expansion in Beijing. c The empirical dynamic of land use change. d Introductions of the proposed relationship, illustrating how grain production loss correlates with UGS coverage

Full size image

In this study, UGS expansion refers to the process of creating parks, gardens, green corridors, or other forms of vegetation in newly developed urban areas, thereby increasing the extent and coverage of urban green spaces within a city. Utilizing Landsat images, we confirmed that UGS expansion was responsible for cropland loss in China during 2000‒2020 (Fig. 2b, c). We introduced concept curves depicting grain production loss (due to cropland loss) contributing to UGS growth (Fig. 2d). As UGS is part of the urban expanded area, the loss of grain production may not fully contribute to UGS growth (i.e., the straight line), and plausible relationships are represented by convex or concave curves. A concave curve implies the existence of a saturation point (Fig. 2d), indicating that the contribution becomes ineffective after a certain threshold. Conversely, a convex curve shows a cumulative threshold point (Fig. 2d), signifying that the contribution becomes effective when the loss of grain production accumulates to a certain extent. These concept curves unveil the cost of grain production on UGS growth and help quantify the optimizing threshold.

Data and preprocessing

The spatiotemporal variations in UGS were analyzed using the Global Urban Land Use/cover Composites (GULUC-30) dataset, the sole dataset providing global UGS fraction at the pixel scale for 2000, 2010, and 2020. This dataset is robust, with an overall accuracy of urban boundaries exceeding 91% and a root-mean-square error of UGS coverage at 13% (National Remote Sensing Center of China 2020). A notable advantage of GULUC-30 lies in its globally standardized definitions of urban boundaries and UGS, enabling meaningful comparisons between cities across the world. The urban boundary denotes specific areas characterized by concentrated clusters of buildings, roads, and corresponding infrastructure, with a significantly higher urbanized population rate than surrounding areas. UGS within the urban boundary encompass natural or artificial vegetation such as parks, gardens, and urban forests. The dataset is accessible through the ChinaGEOSS Data Sharing Network at a resolution of 250 m.

To minimize uncertainties in quantifying cropland loss, we employed three land use/cover products: China Land Use/Cover Dataset, China Land Cover Dataset, and China's Annual Cropland Dataset (Fig. S1) (Kuang et al. 2022b; Tu et al. 2023b; Yang and Huang 2021). Despite variations in mapping schemes, all three products demonstrated robust accuracy. We focused on the “cropland” category from 2000 to 2020, with three products upscaled to a resolution of 250 m, aligning with GULUC-30.

Annual grain production data at the provincial scale from 2000 to 2020 were obtained from the National Bureau of China (NBC). Grain yield (in t/ha) at the provincial scale was computed by dividing grain production by cropland area (Fig. 3) (Kuang et al. 2022a). Historic grain yields for the periods 2000–2010 and 2010–2020 were averaged to establish baseline figures. Scenario-utilized grain yield represents the average for 2015–2020. Following NBC's grain classification, we computed the yield of rice, wheat, maize, beans, and tubers, estimating their respective production losses. The use of three cropland products allowed us to generate three estimations of grain production loss, with the average ± standard deviation reported as the final result.

Fig. 3
figure 3

Technical route and research contents of this study

Full size image

Auxiliary data included urban population for 2000–2020 and future estimations under the Shared Socioeconomic Pathways (SSPs). Historic urban population data were sourced from WorldPop (worldpop.org), and future estimations were derived from open-access SSPs products (Chen et al. 2020b; Wang et al. 2022b). Additionally, we quantified UGS coverage across high-income nations, based on the 2023 World Bank category (Dong et al. 2022), for designing scenarios (Fig. 3).

Indicators and spatiotemporal analysis

UGS change

Urban expansion quantifies the changes in urban areas from 2000 to 2010 and 2010 to 2020 within the 3402 investigated urban areas. We assessed three UGS change indicators—alterations in UGS area (ΔUGS), UGS coverage, and per capita UGS area (UGSpc) across the study periods. The coverage indicator provides a comprehensive view of the abundance of UGS, while the UGSpc metric offers an overall insight into individual accessibility to UGS, also referred to as UGS exposure. For each urban areas, UGS coverage is the ratio of the UGS area to the urban area, and UGSpc is calculated as the UGS area divided by the urban population.

Cropland loss to UGS (ΔCroplandUGS)

Employing spatial analysis methods, we determined the spatial extent of urban expansion from 2000 to 2010 and 2010 to 2020 at the patch scale. This extent was then used to mask UGS at the end of each period, representing ΔUGS for that duration. Overlaying ΔUGS with cropland pixels at the initial time of each period enabled the quantification of cropland loss to UGS.

Grain production loss due to ΔCroplandUGS (ΔGrainUGS)

Through spatial analysis, we allocated each urban patch to its corresponding province. We utilized estimated grain yield and the spatial extent of ΔCroplandUGS to quantify grain production loss within a given period.

Curve fitting for grain production loss versus UGS coverage

We tested the proposed concept curves depicted in Fig. 2d, confirming the relationship of ΔGrainUGS versus UGS coverage. This relationship is shaped by a concave curve or, less commonly, a straight line. The linear relationship, when present, is fitted using simple linear regression. The concave curve is regressed by the exponential decay method at the 0.001 significance level, with the equation:

$$\text{y = }\sum_{\text{i=1}}^{\text{n}}{{\text{A}}}_{\text{i}}{{\text{e}}}^{-\left(\frac{\text{x}}{{\text{C}}_{\text{i}}}\right)}\text{+}{\text{y}}_{0},$$
(1)

where y and x represent UGS coverage and ΔGrainUGS, respectively. N is the phase (1, 2, or 3), and other parameters include the constant Ci, the amplitude Ai, and the offset y0. We subsequently determined the saturation point of concave curves, as illustrated in Fig. 2d, by iteratively sliding each point and identifying the position exhibiting the most significant variance gaps before and after the intervals (considering three points). The focus is on 90% of the maximum UGS coverage (Dong et al. 2020).

Scenario setting and modeling

We envisaged three scenarios outlining the future of urban greening in China, denoted as S30%, SHIC, and SBAU. The S30% scenario is characterized by achieving a 30% UGS coverage across all urban areas, aligning with the 3-30-300 initiative (Browning et al. 2024). The SHIC scenario aims to synchronize national UGS coverage with the observed level in high-income countries as of 2020 (estimated at a mean of 37% from GULUC-30). Lastly, the SBAU scenario postulates that China can maintain its UGSpc at the 2020 level even as the urban population peaks in the future. In other words, Chinese urbanites can have sustained access to UGS and its benefits despite the rapid pace of future urbanization. We selected 2040 as the projected peak year for China’s urban population, estimated to reach approximately 1007 million. Additionally, UGSpc in scenarios S30% and SHIC was quantified using the 2040 population.

The modeling schemes employed varied across scenarios. We employed the Random Forest algorithm, integrating spatially explicit drivers such as population, gross domestic product, road map, and digital elevation model, to generate a raster representing the urban expansion potential (Fig. S1) (Kuang et al. 2022a). This raster provides a continuous value indicating the likelihood of urban expansion. Subsequently, the future ΔCroplandUGS was estimated by masking cropland in 2020 with the expanded urban area from 2020 to the respective scenario. In the case of scenarios S30% and SBAU, a bottom-up modeling approach was applied. This involved evaluating the demand for ΔUGS and subsequently modeling urban expansion for each urban areas requiring urban greening. Conversely, modeling the SHIC scenario entailed assessing the gap in UGS coverage between China and high-income countries and generating urban extent at the national scale based on the expansion potential data.

Results

Observed cropland loss to urban green space across China

The accelerated loss of cropland was observed in UGS in China from 2000 to 2020. The ΔUGS, representing the growth of UGS area, increased by 1.6-fold from 2000‒2010 (0.78 Mha) to 2010‒2020 (1.24 Mha) in China, constituting 30.55% and 32.10% of the expanded urban area during these respective periods (Fig. 4a). The national UGS area quadrupled from 0.68 Mha to 2.71 Mha between 2000 and 2020. Among geographic regions, NC experienced the maximum growth in UGS coverage (7.73%) from 2000 to 2020 (Fig. 4a). Following closely are EC (7.06%) and CSC (6.78%), even though the largest ΔUGS was found in EC, followed by CSC and NC (Fig. 4a, b).

Fig. 4
figure 4

Patterns of UGS growth and its associated cropland change in China, 2000‒2020. In panels a, c, e, ∆UGS denotes the area of UGS growth; ∆CroplandUGS denotes the lost area of cropland to UGS; ∆Croplandurban denotes the lost area of cropland to urban areas. Panels b, d For visualization purposes, spatial changes have been aggregated into grids at a resolution of 30 km. Whiskers indicate standard deviations

Full size image

Correspondingly, the ΔCroplandUGS increased by 1.3-fold from 2000‒2010 (0.51 ± 0.11 Mha) to 2010‒2020 (0.67 ± 0.22 Mha) (Fig. 4c). Aligned with the geographic patterns of ΔUGS, the maximum ΔCroplandUGS is observed in EC (0.52 ± 0.13 Mha) from 2000 to 2020, followed by CSC (0.34 ± 0.10 Mha) and NC (0.14 ± 0.06 Mha), and least in NWC (0.04 ± 0.01 Mha) (Fig. 4d, e). In particular, NWC is the sole region where ΔCroplandUGS decreased from 2000‒2010 to 2010‒2020 (Fig. 4e), largely because of the weakened urban expansion (Fig. S2).

Furthermore, the ΔCroplandurban (cropland contribution to urban expansion) in China has significantly decreased from 55.86 ± 6.26% to 40.89 ± 5.53% between 2000‒2010 and 2010‒2020 (Fig. S2). In contrast, the proportion of ΔCroplandUGS to ΔCroplandurban has risen from 2000‒2010 (35.44 ± 5.94%) to 2010‒2020 (41.10 ± 9.55%). The proportion is 38.47 ± 8.79% across 2000‒2020. Hence, while the pressing concern of cropland loss to urban expansion has been mitigated, greater attention is warranted for the direct encroachment of ΔUGS on cropland.

Nonetheless, the proportion of ΔCroplandUGS to ΔCroplandurban shows approximate geographic patterns. Across 2000‒2010, the maximum proportion (i.e., NC) and the minimum (i.e., EC) showed a numerical difference at 9.73% (Fig. 4f). Despite the widening difference to 13.20% between 2010 and 2020, the primary regions of ΔCroplandUGS (i.e., EC, CSC, NC) maintain a maximum-minimum difference at 6.77%, similar to the period 2000‒2010 (Fig. 4f). Moreover, those proportions are alike across five urban size groups, both 2000‒2010 and 2010‒2020 (Fig. 4g). Hence, the ΔCroplandUGS is a national issue lacking significant regional differences and compelling optimization cases. The shift in China's urban expansion from extensive sprawl to the growth of smaller areas suggests the need to optimize ΔCroplandUGS for smaller urban areas (Fig. 4g) (Kuang 2020).

Amplified undermining of grain production from 2000 to 2020

As anticipated, the impact of ΔCroplandUGS on grain production across China has intensified from 2000 to 2020. The ΔGrainUGS amounted to 1.61 ± 0.34 million t and 2.60 ± 0.83 million t in China during 2000‒2010 and 2010‒2020, respectively (Fig. 5a). In conjunction with the exacerbated ΔCroplandUGS (Fig. 4c), the increase in grain yield from 2000‒2010 to 2010‒2020 contributed to a 21% enlargement in ΔGrainUGS (Fig. S3).

Fig. 5
figure 5

Patterns of grain production loss from cropland transforming to UGS, 2000‒2020. Panel a illustrates the temporal change of national ΔGrainUGS, while panels c, d, g display the variation of ΔGrainUGS across different regions, urban size groups, and grain types, respectively. Panels b, f, g For visualization purposes, spatial changes have been aggregated into 30 km grids. Whiskers indicate standard deviations

Full size image

The spatial pattern of ΔGrainUGS mirrored that of ΔCroplandUGS (Fig. 5b, c). The most substantial ΔGrainUGS occurred in EC, rising from 0.85 ± 0.18 million t to 1.18 ± 0.35 million t from 2000‒2010 to 2010‒2020 (Fig. 5c). Following closely were CSC (0.47 ± 0.09 million t to 0.90 ± 0.30 million t) and NC (0.12 ± 0.05 million t to 0.24 ± 0.10 million t), with the least in NWC (0.03 ± 0.01 million t at both periods) (Fig. 5d, e). Notably, from 2000‒2010 to 2010‒2020, the ΔGrainUGS in EC increased by 1.4-fold, ranking behind NEC (3.7-fold), NC (twofold), and CSC (1.9-fold) (Fig. 5c). Additionally, significant ΔGrainUGS occurred in I-type urban areas, followed by III-type areas (Fig. 5d). Thus, ΔGrainUGS hotspots encompass EC and CSC cities with urban populations under 0.5 million and those between 1 to 5 million from 2000 to 2020, emphasizing the need for heightened concerns in NEC and NC with a substantial increment of ΔGrainUGS.

The production of rice, a staple in China, has been notably impacted by ΔCroplandUGS. The loss in rice production was 0.76 ± 0.14 million t from 2000 to 2010, representing a twofold loss compared to wheat (0.37 ± 0.10 million t) and corn (0.37 ± 0.08 million t) (Fig. 5e). Although this gap narrowed from 2010 to 2020, the production losses of rice (1.08 ± 0.33 million t), wheat (0.71 ± 0.23 million t), and corn (0.58 ± 0.20 million t) amplified (Fig. 5e).

Loss hotspots have been identified (Fig. 5f, g). Primary losses in rice production occurred in EC, with the share of the national total increasing from 52 to 57% from 2000‒2010 to 2010‒2020, followed by CSC with a decrease from 40 to 34%. Losses in wheat and corn primarily occurred in EC, CSC, and NC (Fig. 5f, g). A notable shift in wheat loss hotspots was observed from EC (with the share decreasing from 66 to 43%) to CSC (with the share increasing from 19 to 41%) from 2000‒2010 to 2010‒2020, while the share in NC remained stable (11% to 13%). Regarding corn loss, the distribution from 2000 to 2010 peaked in EC at 43%, followed by NC (21%) and CSC (17%). However, these shares were redistributed across regions from 2010 to 2020, with values of 30%, 29%, and 22% for EC, CSC, and NC, respectively (Fig. 5f, g).

The saturation effect of ΔGrainUGS has been verified (Fig. 6). Concave curves were fitted at the national scale across the investigated periods (p < 0.001), indicating that the contribution of ΔGrainUGS to UGS coverage becomes inefficient after reaching a saturation point across China (Fig. 6a). The saturation points for ΔGrainUGS shifted from 19.81 × 103 t to 29.06 × 103 t from 2000‒2010 to 2010‒2020. Consequently, approximately 212 × 103 t and 183 × 103 t of ΔGrainUGS were unnecessary in promoting UGS growth in China during those periods. These unnecessary ΔGrainUGS are attributed to the heterogeneous urban size. Urban population is the primary driver of urban expansions (Mahtta et al. 2022). Consequently, as urban size increases, the loss of grain production rises due to ΔCroplandUGS (Fig. 6b).

Fig. 6
figure 6

Concave relationship between grain production and the UGS coverage. Fitting curves or lines are at a 0.001 significance level in panels c‒h. Panels a, b represent China; panels c‒e display EC, SWC, and NWC, respectively; panels f‒h display I-type, IV-type, and V-type urban areas, respectively

Full size image

Concave curves have been observed in geographic regions (p < 0.001), such as EC and SWC (Fig. 6c, d). Notably, saturation points shifted inversely in EC and SWC from 2000‒2010 to 2010‒2020, and NWC rejected the proposed curves from 2010 to 2020 (Fig. 6c‒e). The backward shift of saturation points in SWC and the absence of a concave curve in NWC were due to insufficient ΔUGS and even the decrease in UGS coverage in NWC from 2010 to 2020.

However, concave relationships or saturation points did not universally exist across urban-size groups. Only the I-type urban exhibits concave curves and saturation points at both investigated periods (p < 0.001, Fig. 6f). Other sizes exhibit either concave curves (p < 0.001) without a saturation point or a linear relationship (p < 0.001) between grain production loss and UGS coverage (Fig. S4, Fig. 6g‒h). For example, ΔGrainUGS directly contributed to UGS growth in V-type urban areas from 2000 to 2020 and IV-type urban areas from 2010 to 2020 (Fig. 6g‒h). Therefore, tailored greening goals should be devised based on varying urban populations to optimize the cost of grain production for UGS growth.

Future urban greening schemes and their impact on grain production

The proposed scenarios anticipate trade-offs between ΔUGS and grain production in the future. The projected ΔUGS to achieve SHIC is the largest at 1.35 Mha in China, nearly 3.3 times and 1.8 times the ΔUGS for realizing S30% (0.41 Mha) and SBAU (0.75 Mha), respectively (Fig. 7a). Furthermore, SHIC is projected to result in upper UGS coverage (37.17%) and UGSpc (40.35 m2) compared to other scenarios at the national scale (Fig. 7a). In particular, for S30%, the scenario with the smallest ΔUGS, the change in UGSpc (− 0.46 m2) from 2020 was significantly distinct from those in SHIC (+ 9.72 m2) and SBAU (+ 3.71 m2) in China. What is more, S30% presents a unique scenario where projected UGSpc is lower than the 2020 level in regions with leading ΔUGS from 2000 to 2020 (i.e., EC, CSC, and NC) (Fig. 7b).

Fig. 7
figure 7

Patterns of UGS growth and its associated grain production loss in proposed scenarios. Panel a depicts the projected UGS area, its coverage, and UGS per capita at the national scale, while panel b presents the same information at the regional scale. Corresponding losses in grain production at the national scale are shown in panel c. Panels d‒f: For visualization purposes, spatial changes have been aggregated into grids at a resolution of 30 km. Whiskers indicate standard deviations

Full size image

Nonetheless, less ΔUGS facilitates safeguarding cropland and grain production. Substantial ΔUGS is expected to be notably practical for promoting UGSpc or UGS coverage in regions where the urban population is slightly increasing (e.g., NEC) or UGS storage is limited (e.g., NWC), as opposed to regions attracting people relocations with ample greenery (e.g., EC and CSC) (Fig. 7b). On the flip side, projected ΔCroplandUGS in SHIC (0.19 ± 0.01 Mha) is 2.8-fold and 1.9-fold greater than S30% (0.07 ± 0.00 Mha) and SBAU (0.10 ± 0.00 Mha) across China despite being behind the observation in 2010‒2020 (Figs. 4e, 7b).

Compared to 2010‒2020, projected ΔGrainUGS has significantly decreased in China, with the maximum in SHIC (0.59 ± 0.02 million t), followed by SBAU (0.30 ± 0.01 million t) and S30% (0.21 ± 0.01 million t) (Fig. 7c). Moreover, the spatial patterns of projected ΔGrainUGS would be distinct to 2010‒2020. EC is the region of projected maximum for S30% and SHIC, while it is second, behind CSC, for SBAU (Fig. 7d‒f). This difference in regional maximum across scenarios is attributed to the potential massive growth of the urban population in CSC, where achieving SBAU requires extensive UGS expansion (Fig. S5). Conversely, the share of ΔGrainUGS in NEC to the nation would be the least in SBAU (1.39%) compared to other scenarios, owing to its slight urban population growth and thus the least projected ΔCroplandUGS (Fig. 7d‒f, Fig. S5).

Grain types experiencing substantial losses across scenarios include rice, wheat, and corn, consistent with 2010‒2020 (Figs. 5f, g, 7d‒f). However, the spatial distribution of grain types' production loss may vary across scenarios due to regional features. In the primary rice production zone (NEC), characterized by minimal UGS coverage in 2020 and a future urban population compared to other regions, achieving a UGS coverage of 30% (S30%) would require a higher increase in ΔUGS (Fig. 7b). This, in turn, significantly elevates the nation's contribution of rice loss to the total ΔGrainUGS (46.07%), surpassing other scenarios. This share would be the minimum in SBAU (40.21%) due to the limited utility of massive ΔUGS in NEC under minimal projected urban population growth. Additionally, the nation's share of wheat loss to total ΔGrainUGS would be the maximum in SBAU (19.34%) due to the potentially massive growth of the urban population in the primary wheat production zone (i.e., the transboundary area of NC, EC, and CSC). In total, the impact of future ΔUGS on grain production is complex, influenced by the greening target of various schemes, regional UGS storage, and regional features in agriculture.

The saturation effect is expected across various greening schemes. Nationally, the saturation points of ΔGrainUGS gradually shift along S30% (6.3 × 103 t), SBAU (7.5 × 103 t), and SHIC (16.8 × 103 t) (Fig. 8a). For these scenarios, the excess ΔGrainUGS is 39.54 × 103 t, 49.42 × 103 t, and 58.59 × 103 t, respectively, to promote UGS coverage nationwide. However, these distinctions in saturation points do not necessarily imply the feasibility of a scenario for balancing ΔUGS and ΔGrainUGS at the national scale due to the future concave relationship, which exhibits substantial heterogeneity across geographic regions and urban sizes (Fig. 8b‒d). All geographic regions generally exhibit concave curves (p < 0.001), with uncertain existence of saturation points, e.g., absence of saturation point in SBAU of NEC and in S30% of SWC (Fig. 8c, d). Additionally, the concave curve is observed across all-size urban areas (p < 0.001), with the absence of saturation points noted in SBAU for III-type urban areas, SHIC for IV-type urban areas, and all scenarios for V-type urban areas (Fig. S6, Fig. 8e‒g).

Fig. 8
figure 8

Concave relationship between grain production loss and UGS coverage in proposed scenarios. Fitting curves are at a 0.001 significance level in panels b‒g. Panel a represents China; panels b‒d display EC, NEC, and SWC, respectively; panels e‒g display I-type, IV-type, and V-type urban areas, respectively

Full size image

Optimizing the future balance between ΔUGS and ΔGrainUGS requires careful consideration of UGS storage and urban population growth. At the national scale, S30% stands out as the most favorable pathway, demanding the minor saturation of ΔGrainUGS compared to other scenarios. However, SBAU is a viable option for NEC due to its minimal need for future urban populations. Especially for urban areas with more than 5 million dwellers, S30% is anticipated to be the vital way to improve UGS coverage to the recommended level with the least costs on grain production (Fig. 8f, g). As a result, the optimal target for UGS coverage in China by 2040 is set at 30.87%, minimizing cropland loss (0.06 Mha) and grain production loss (0.18 million t).

Discussion

Critical but concealed impact of UGS expansion on cropland

The issue of cropland loss to urbanization is a global challenge, posing significant threats to agricultural sustainability. China has contributed a quarter of the global amount of cropland lost to urbanization from the 1990s to the mid-2010s, highlighting an urgent need to optimize the competition for land between urbanization and agriculture (Huang et al. 2020; van Vliet et al. 2017). Previous studies have reported varying estimates of China's cropland loss to urbanization, ranging from 3.22 Mha (2003‒2016) and 4.22 Mha (1993‒2016) to as high as 12 Mha (1990‒2019) (He et al. 2017; Qiu et al. 2020; Tu et al. 2023a). To mitigate uncertainties, we employed three land use cover products, resulting in a more constrained estimate of 3.03 ± 0.45 Mha for the period 2000‒2020. This closely aligns with a study focused on 2003‒2016 (Qiu et al. 2020), providing a robust foundation for a new understanding of the cropland loss owing to UGS expansions in the urbanization context. These results are valuable for decision-makers grappling with the need to optimize competition for land between urbanization and agriculture.

The joint management of urban human well-being alongside productive and natural landscapes is essential for advancing landscape sustainability and achieving the multifaceted targets of the Sustainable Development Goals (SDGs) (Wu 2019). Our study indicates that the area of cropland loss to UGS expansions in China accounted for 38.47 ± 8.79% of the cropland loss to urbanization from 2000 to 2020, emphasizing the significant role of UGS expansion in urbanization-induced cropland consumption. Notably, ~ 4.2 ± 1.4 million t of grain production were consumed by the observed cropland loss to UGS expansions during this period, equivalent to the annual food intake of around 9.3 million people in China (based on the nation's 2015 standard of per capita 450 kg/year). While UGS expansion contributes to human well-being by fostering ecosystem services in built-up environments, it simultaneously depletes the provisioning services of agricultural ecosystems, particularly in smaller-sized urban areas with populations of less than 0.5 million (Fig. 5d). This depletion is expected to manifest in developing countries soon as urbanization accelerates (Bren d’Amour et al. 2017; Chen et al. 2020a; Tu et al. 2023a), accompanied by a growing demand for ecosystem services from UGS expansions. Consequently, it is imperative to regulate massive urban surface greening to mitigate its impact on regional and national food security, especially for nations with poor agricultural resources and management.

Saturation effect navigates future urban greening scheme

The proposed saturation effect offers a means to strike a balance between enhancing urban human well-being and preserving natural and productive landscapes. The loss in grain production contributes to UGS coverage, following a concave trajectory with a saturation point (Fig. 6). In other words, while urbanization consumes productive landscape and its associated benefits, only a fraction of these resources can effectively contribute to UGS growth, thereby improving the sustainability, resilience, and livability of cities. In China, approximately 87% to 93% of the grain production loss owing to cropland loss to UGS expansions can be considered effective from 2000‒2010 to 2010‒2020, suggesting that the loss of 395 × 103 t of grain production could potentially have been avoided during these two decades (Fig. 6a).

The identified saturation point offers valuable insights into preserving natural and productive landscapes when pursuing desired ecosystem services through UGS expansions. For instance, induced by changes in land use, the reduction in ecosystem carbon storage and biodiversity loss contribute to UGS and its ecosystem services, including cooling, air purification, and biodiversity maintenance for built-up areas (Bren d’Amour et al. 2017; Endreny 2018; Seto et al. 2012; Wang et al. 2022a). It is crucial to acknowledge that UGS expansions, akin to urban expansions, may have influenced ecosystem services and biogeophysical processes in distant regions, both through cropland redistribution and landscape homogenization (Kuang et al. 2022a; Lemoine-Rodríguez et al. 2020). The suggested saturation effect holds promise for unraveling the trade-offs among SDGs like Sustainable Cities and Communities (SDG11) versus Zero Hunger (SDG2), Life on Land (SDG15), and Climate Action (SDG13), as well as for identifying strategies to mitigate those trade-offs.

We have identified a scheme to balance UGS expansions and minimize UGS expansion-induced grain production loss in China. At the national scale, strategically selecting S30% scenario promotes urban well-being through UGS expansion while incurring minimal cropland and grain production loss. On a regional scale, the future demand for UGS expansions depends on shifts in urban population, making SBAU a prudent choice for regions like NEC with lower urban population growth. It is also suggested to pay attention to regulating trade-offs between UGS expansion and cropland protection in underdeveloped cities with less urban population. Consequently, we advocate for improving China’s UGS coverage by 2.57% from 28.30% to 30.87% between 2020 and 2040. This is crucial for meeting the growing demand for ecosystem services from UGS as well as cropland protection in response to future urbanization trends. In particular, identifying a saturation threshold is critical for advancing SDGs in developing countries challenged by unequal UGS exposure, poor self-sufficiency in grain production, and growing demand for improving human well-being in cities and food security.

Implications for expanding UGS in compact cities

While urban densification offers promise in controlling urban expansion and preserving regional ecosystems, the increase in population density exacerbates unequal access to UGS and its services. For example, UGS in China, replaced by impervious materials with urban population growth, saw a yearly increase of 423 million urbanites without access to cooling services from 2003 to 2015 (Dong et al. 2022; Zhang et al. 2023). Although the urban densification framework allows the creation of mini UGS with multifunctional uses or building planting, larger UGS are more effective in providing ecosystem services. Thus, cities face the dilemma of choosing between UGS expansion and maintaining urban compactness, particularly amid the growing impact of climate change on human well-being.

This study underscores that a scheme involving massive UGS expansion (e.g., scenarios SHIC) could be detrimental to China, facing challenges related to food security, cropland loss, and population growth. For instance, the doubling of grain production loss is projected (Fig. 7c) with only a 5.6% increment in greening targets shifting from S30% to SHIC across China. We argue that slight UGS expansion is compatible with the urban densification framework, emphasizing the importance of careful greening targets to enhance human well-being in urban areas. Promoting urban agriculture in China, including farms and vegetable gardens in peri-urban and fragmented urban areas during UGS expansion, not only alleviates pressure on the urban food supply chain but also enhances urban sustainability (Zhu et al. 2024).

Concerning the S30% scenario, in addition to minimized losses in cropland and grain production, China’s per capita UGS area by 2040 is estimated to approximate the level of 2020 (Fig. 7a). Therefore, the national scheme of 30.87% UGS coverage proposed by this study suggests a fundamental maintenance of urbanites' accessibility to UGS and its services in the future. It is noted that scenario projections indicate a reduction in future amount of cropland loss to UGS expansions compared to 2010‒2020, even in the SHIC scenario. Consequently, recognizing that future UGS and urban expansions may predominantly consume other ecosystems in addition to agriculture is crucial for urban planning, aiming for harmony between human well-being in urban areas and regional sustainability. However, it is necessary to acknowledge that this study solely focuses on one aspect of UGS expansion—its impact on cropland stock and grain production. Future research should assess the multifaceted consequences of UGS expansions on natural and productive landscapes, providing informed strategies for addressing these trade-offs comprehensively.

Conclusions

This proof-of-concept study underscores the concealed impact of urban greening on cropland and grain production in China from 2000 to 2020, attributing nearly 40% of cropland loss to UGS expansions. Despite introducing a trade-off with urban densification, UGS expansion emerges as a direct and effective approach, enhancing urbanites' access to nature and improving human well-being. In cities grappling with unequal and insufficient UGS exposure due to population growth and urban densification, an informed UGS expansion strategy is deemed essential. The proposed saturation effect serves as a valuable tool for decision-makers and researchers aiming to transform the original ecosystem and its services into desired urban greening and associated benefits. As demonstrated in our study, setting a specific UGS coverage target of 30.87% until 2040 enables China to maintain UGS exposure with minimized cropland loss, even amid population growth. We advocate for a coordinated approach involving UGS expansion and cropland protection measures to advance sustainable urban development and regional food security.