Introduction

Global flood risk has been increasing over the past decades (Bradshaw et al. 2007; Muis et al. 2015). Hereby, flood risk is defined as a combination of hazards, exposure and vulnerability (Koks et al. 2015), where escalating exposure may be the main driver of growing risk (Bubeck et al. 2011; Visser et al. 2014). Low-lying coastal cities are at great risk of extreme flooding (Erban et al. 2014; Jongman et al. 2014; Scussolini et al. 2017; Shan et al. 2021), and consequently of higher flood exposure as climate change and rapid urbanization (Jongman et al. 2012; Willner et al. 2018). Understanding how climate change and urbanization affect and contribute to changing flood exposure is critical to advancing the sustainable development goals outlined by the United Nations (Güneralp and Seto 2013; Winsemius et al. 2015; Popp et al. 2017; Reichstein et al. 2021).

Urban expansion refers to the process of spatial expansion of urban scale due to changes in land use, and this process is accompanied by landscape changes in land use, land cover and ecosystems(Zhao et al. 2015; Song et al. 2022). On the one hand, urban expansion along with socio-economic development can provide better healthcare, education and cultural resources which improve the welfare of its people(Wu 2014; Sun et al. 2018). On the other hand, urban expansion can also bring a series of negative environmental consequences that have an influence on the sustainability of the region(Wu 2013; Peng et al. 2021). Therefore, the process of urban expansion is closely linked to global and regional sustainability, and related research is of great significance in maintaining and promoting sustainable development(Brandt et al. 2012; Wu 2021).

China is among the most seriously affected countries by floods (Liu et al. 2013; Du et al. 2019; IPCC 2022). During 2000–2020, China experienced 211 floods, resulting in 12,922 causalities and a cumulative loss of US $268.64 billion (EM-DAT 2023). Concurrently, the coastal population had grown by about 90 million, boosting the urbanization rate from 37.44 to 64.87%, and enlarging urban lands from 6.68 × 104 to 1.25 × 105 km2 (NBS 2022). The flood impacts facing coastal China would further exacerbate against the background of climate change (Fang et al. 2017; Wang et al. 2017; Gao et al. 2019; O'Donnell 2019), as 5.7%–6.0% of China’s coastal areas would be drowned by 2050 in the event of a 1000-year coastal flood (currently 5.3%) (Wang 2022). Land use adjustment and optimization is one of the main approaches to flood adaptation (Du et al. 2021; Duijndam et al. 2022), which can aid decision-makers in formulating effective adaptation or mitigation measures (Liu et al. 2017; Verburg et al. 2019; Johnson et al. 2021). Consequently, the optimization of urban land in coastal China carries significant implications for adapting to flood risk (Kim and Rowe 2013).

Studies had been conducted to evaluate the Urban Exposure to Flood (UEF) in coastal areas of China at various scales. Du et al. (2018) assessed the distribution and expansion of urban land in floodplain in China from 1992 to 2015, and found that 68.93% of exposed urban lands were clustered in coastal basins, and 68.22% of the newly increased was located there. Fang et al. (2021) revealed that the nighttime light intensity in coastal regions, such as Beijing-Tianjin-Hebei, Yangtze River Delta, and Pearl River Delta, had significantly expanded and increased during 1992–2020, with elevated flooding exposure and faster growth rate. Wu et al. (2022) investigated the flood exposure of urban land in China coastal flood-prone areas from 2000–2020, and discovered that areas with low flood protection levels had an average annual growth rate of urban land as high as 9.40%, 1.44 times higher than that of the entire study area. While these studies did not address the future, where the tension between urbanization and flooding would become more pronounced. To ensure the sustainable development of coastal cities, it is crucial to understand the spatial and temporal dynamics of flood exposure, as well as the future trends under changing environmental conditions (Jongman et al. 2014; Jongman 2018; Willner et al. 2018).

Case studies on future urban land expansion in China’s coastal floodplains were also available for localities. Lin et al. (2020) and Sun et al. (2022) conducted scenario-based coastal flood exposure assessments using the future land use simulation (FLUS) model for the Guangzhou metropolitan area and Shanghai, respectively, and found that without proper urban planning, more built-up areas in both regions would be exposed to flooding. Xu et al. (2021a) established different urban growth scenarios based on the CA–Markov model and combined them with the sea level rise scenarios to dynamically evaluate the flood exposure of buildings and infrastructure in Xiamen in 2050, and revealed that sea level rise and urban growth would increase the overall flood exposure to varying degrees. Zhao et al. (2023a) constructed a method combining Bayesian network and patch-generating land use simulation (PLUS) model to predict the future flood risk in Zhengzhou City under various scenarios in 2035, and discovered that there was a continuous growth in exposure index, which indicated that the flood exposure was increasing. Despite these specific studies on cities, there remains a lack of research on future UEF in coastal China, and the relative importance of flood variation due to climate change and urban growth on future changes in coastal flood exposure has not been quantified. A more comprehensive and holistic approach is therefore urgently needed to assess the complexity of multiple possible scenarios of urbanization and dynamic flood exposure.

To fill the research gap, our objective is to analyze the discrepancy of UEF under multiple climate change and flood-risk based planning scenarios. Furthermore, we endeavor to assess the effectiveness of various urban planning strategies in reducing flood exposure and determining the relative contributions of climate change and urban growth on UEF. By doing so, we aim to explore potential pathways for future flood adaptation in coastal China. To commence our analysis, we simulated the urban sprawl pattern expected by 2050 using the FLUS model under various urban planning and climate change scenarios specifically tailored to the coastal areas of China. Subsequently, we overlaid the multi-period flood inundation datasets to diagnose the discrepancies of UEF under different scenarios. Additionally, we quantified the respective contributions of urban growth and flood variation to future UEF dynamics. This comprehensive analysis holds immense significance in seeking feasible mitigation strategies at a regional level and enhancing the land system resilience against flooding.

Study area and data

We analyzed the variation in UEF at both regional and provincial scales. The study area encompassed 15 provincial units, such as Tianjin, Jiangsu, Shanghai, Zhejiang, and Guangdong. Due to limited data availability, Hong Kong, Macao, and Taiwan were not included in this study (Fig. 1). The administrative boundaries were obtained from the Resource and Environment Science and Data Center (https://www.resdc.cn/).

Fig. 1
figure 1

Study area and its land use pattern in 2020

Full size image

The urban land from 2000 to 2020 were obtained from He et al. (2022b), with a spatial resolution of 1km, a kappa coefficient of 0.60, and an overall accuracy of 92.62%. This dataset could provide data support for simulating future urban expansion in coastal China. Historical urban population data were derived from the China Statistical Yearbook (http://www.stats.gov.cn/sj/ndsj/), and future urban population dataset under different scenarios were retrieved from Chen et al. (2020b). For the flood inundation maps, we utilized the work of Wang (2022) for the years 2020 to 2050. Specifically, we selected coastal inundation data corresponding to RCP4.5 and RCP8.5 for eight different return periods (2, 10, 25, 50, 100, 250, 500, and 1000 years) in 2050, representing sea-level rise scenarios of 5%, 50% and 95%, respectively. To delineate the spatial extent of the wetlands protection scenario, we employed the wetland dataset by Mao et al. (2020), resampling the original data with a 30 m resolution to 1km. Additionally, the Digital Elevation Model (DEM), Gross Domestic Product (GDP), and Normalized Difference Vegetation Index (NDVI) datasets were all acquired from The Resource and Environment Science and Data Center (http://www.resdc.cn), where NDVI was used to extract the spatially constrained extent of ecological protection scenario within a certain threshold. Geographic ancillary data include road networks, city sites, coastlines, rivers, etc. (see table S1 for details), were sourced from OpenStreetMap (https://www.openstreetmap.org).

Methods

Defining of urban land flood exposure (UEF)

The UEF in this study was defined as the maximum extent to which urban land was inundated by flood, without considering the in-depth inundation (Du et al. 2018). We analyzed UEF under Representative Concentration Pathways (RCPs), specifically RCP4.5 (a climate scenario with government intervention) and RCP8.5 (a baseline scenario without climate change policy intervention). For each RCP, we considered flood inundation datasets corresponding to eight different return periods.

The average annual change rate (AACR) was used to reflect the dynamics of UEF (Du et al. 2018), which was calculated as follows:

$$AACR=\left(\sqrt[\left(t2-t1\right)]{\frac{{X}_{t2}}{{X}_{t1}}}-1\right)\times 100\text{\%}$$
(1)

where \(AACR\) was the average annual change rate; \({X}_{t1}\) and \({X}_{t2}\) referred to the area of urban lands in years \(t1\) and \(t2\), respectively.

Scenarios setting

Referencing to the ScenarioMIP (Scenario Model Intercomparison Project) and socioeconomic data from the SSPs (Shared Socioeconomic Pathways) (O’Neill et al. (2013), the general framework depicting the scenario-based assessment of future urban land flood exposure was presented in Fig. 2. Limited by the data availability, the SSP2-4.5 and the SSP5-8.5 were chosen. Drawing upon relevant studies (Mustafa et al. 2018a; Liu et al. 2020b; Sun et al. 2022), we designed five urban planning strategies as follows:

  1. i.

    The Business as Usual (BAU) scenario, representing the baseline scenario without any interventions being implemented;

  2. ii.

    The Wetlands Protection (WP) planning, which maintains the same quantities as BAU scenario, but restricting the spread of newly added urban land into wetlands reserves;

  3. iii.

    The Ecological Protection (EP) planning is consistent with BAU in terms of quantities, but imposes limitations on urban growth in areas with high vegetation cover;

  4. iv.

    The Densification Development (DD) planning, which allows for unrestricted spatial expansion but imposes a limit on the amount of future urban land demand;

  5. v.

    The Sustainable Development (SD) planning, which shares the same quantities as DD planning while incorporating spatial constraints from both WP and EP planning.

Fig. 2
figure 2

The framework of scenario-based assessment of future urban land flood exposure

Full size image

Estimation of urban land demand

Following He et al. (2021), a linear regression model was used to estimate the urban land demand under the two SSPs. In this study, population was used as an independent variable to estimate the future urban land demand. Initially, we developed linear regressions for the historical urban population and urban land area of 12 coastal provinces using the pane data in 2000, 2005, 2010, 2015 and 2020. The equation can be expressed as:

$${UA}_{s}^{t}={\beta }_{0}+{\beta }_{1}\times {UP}_{s}^{t}$$
(2)

where \({UA}_{s}^{t}\) and \({UP}_{s}^{t}\) referred to the urban land area and urban population in year \(t\) under scenario \(s\) respectively. \({\beta }_{0}\) and \({\beta }_{1}\) referred to the fitting parameters. The \({R}^{2}\) of the regression equation was 0.92 at a 95% significance level. Upon establishing the regression equation, we estimated future urban land demand under SSP2 and SSP5 by using future urban population data. It should be noted that, although the population is expected to decelerate over time we assumed that once non-urban land is converted to urban land, it cannot be converted to other land use types, following Chen et al. (2020a), in other words, we assumed that the urban land would not shrink accordingly. Within the DD planning, we cut the amount of additional urban land demand by 10%.

Spatial simulation of urban land

The FLUS model was adopted for spatial simulation in this study. It is a land-use change simulation model that combines deep learning with Artificial Neural Network (ANN) and Cellular Automata (CA) (Liu et al. 2017). Specifically, we calculated the probability of non-urban pixels to be converted to urban pixels as follows:

$${\text{TP}}_{p,k}^{t}={sp}_{p,k}\times {\Omega }_{p,k}^{t}\times {Inerria}_{k}^{t}\times (1-{sc}_{c\to k})$$
(3)

where \({\text{TP}}_{p,k}^{t}\) denoted the combined probability of grid cell \(p\) to covert from non-urban land to urban land \(k\) at iteration period \(t\); \({sp}_{p,k}\) denoted the probability-of-occurrence of urban land \(k\) on grid cell \(p\) output by Artificial Neural Network(ANN); \({\Omega }_{p,k}^{t}\) denoted the neighborhood effect of urban land \(k\) on grid cell \(p\) at iteration period \(t\); \({Inerria}_{k}^{t}\) denoted the inertia coefficient of urban land \(k\) at iteration period \(t\); \({sc}_{c\to k}\) denoted the conversion cost from non-urban land \(p\) to urban land \(k\); and \((1-{sc}_{c\to k})\) denoted the ease of conversion. Following Liu et al. (2017), Liao et al. (2020) and He et al. (2021), 11 driving factors, namely DEM, slope, population density, GDP, distance to road networks, cities, coastlines, and rivers were selected (Figure S1).

Accuracy verification

To quantitatively evaluate the simulation results, we constructed confusion matrix for computing the overall accuracy (OA) and Kappa coefficient, and further verified the agreement of the changes using figure of merit (FoM), which were superior in evaluating the accuracy of the simulation (Pontius et al. 2007). The expression of FoM was as follows:

$$FoM=\frac{B}{A+B+C}$$
(4)

where \(A\) was the area of error where the actual observation shifted but predicted as non-shift, \(B\) was the area of correctly predicted shift, \(C\) was the area of error where observed non-shift but predicted as shift.

The Pearson correlation coefficient (Schober et al. 2018) was used to verify the correlation between the future urban area at county scale of this study and existing studies:

$$r=\frac{\sum_{i=1}^{n}({A}_{i}-\overline{A })({B}_{i}-\overline{B })}{\sqrt{\sum_{i=1}^{n}{({A}_{i}-\overline{A })}^{2}}\sqrt{\sum_{i=1}^{n}{({B}_{i}-\overline{B })}^{2}}}$$
(5)

where \(r\) referred to the correlation coefficient; \({A}_{i}\) and \(\overline{A }\) referred to the area and average area of urban land under SSPs of our study, respectively, and \({B}_{i}\) and \(\overline{B }\) referred to the area and average area of urban land of other studies.

Contribution analysis

To evaluate the relative importance of urban growth and flood variation to the dynamics of future UEF, we set three distinct cases:

  1. i.

    Urban Growth Only (UG only), considering only future urban growth while assuming static variation, that is, maintaining the flood inundation extent as observed in 2020;

  2. ii.

    Flood Variation Only (FV only), considering only dynamic flood variation while assuming static urban growth, that is, retaining the urban land pattern as observed in 2020;

  3. iii.

    Urban Growth and Flood Variation (UG + FV), combining both future urban growth and flood variation.

Referring to Vousdoukas et al. (2018) and Xu et al. (2021a), the relative contributions of urban growth and flood variation (climate change) were quantified as follows:

$${RC}_{i}=\frac{{UFE}_{i}}{{UFE}_{Baseline}}-1$$
(6)
$${\rho }_{i}=\frac{{RC}_{i}}{\sum {RC}_{i}}\times 100\%$$
(7)

where \({RC}_{i}\) was the relative change of UEF in the case considering only urban growth or flood variation; \({UEF}_{i}\) was the urban flood exposure in the case considering only urban growth or flood variation; \({UEF}_{Baseline}\) was the UEF in 2020; and \({\rho }_{i}\) was the relative contribution of urban growth or flood variation to the changes of future UEF.

Results

Reliability of the urban expansion simulation

Urban expansion from 2000 to 2010 and 2010 to 2020 were simulated and the results were compared against the actual urban lands in 2010 and 2020, respectively. The calibration of the model for these two simulation periods yielded Kappa coefficients of 0.71 and 0.76, with an OA of 0.99 and 0.98, and FoM of 0.27 and 0.18, respectively. Existing applications of land change modelling usually reported FoM values of 0.1–0.3 (Chen et al. 2020a), thereby suggesting the reliability of our simulations for coastal China.

Following Chen et al. (2020a), we compared the results of the BAU scenario for future urban land in coastal China with those of He et al. (2022c), Li et al. (2021) and Liao et al. (2020), and found that the outcomes were fairly consistent across SSP2 and SSP5. At the county scale, we specifically selected counties with a population of more than 1.5 million in 2020 and compared our data with three other datasets as to whether there was a correlation between the areas. The results showed that all correlation coefficients higher than 0.91 (P < 0.01) (Fig. 3), proving the reliability of our simulation results. Additionally, at the urban scale, we evaluated the performance of our model by comparing the kappa coefficients for each pane under SSP2, with mean kappa of 0.74, 0.65 and 0.73 compared to He et al. (2022c), Li et al. (2021) and Liao et al. (2020), respectively (Fig. 4).

Fig. 3
figure 3

Comparison of urban land area estimation in 2050 with three previous projections

Full size image
Fig. 4
figure 4

Examples of simulated urban land patterns for the period 2000–2020 compared with Landsat images and for the period 2020–2050 under SSP2 compared with three previous projections

Full size image

The UEF under BAU scenario during 2020–2050

In 2020, the UEF in coastal China ranged from 3494 to 9879 km2 for 2-year to 1000-year flood, accounting for 9.19%–14.06% of the total inundated area, with an average annual change rate of 6.64%–8.55% during 2000–2020, which was 1.28–1.61 times higher than that of non-exposed urban lands (5.22%–5.30%). For 1000-year event, the majority of the UEF (9887 km2, 91.07%) was concentrated in Zhejiang, Guangdong, Shanghai, Jiangsu, and Tianjin (Table S2). Among these provinces, Zhejiang and Guangdong exhibited the highest UEF, covering an area of 2167 km2, followed by Shanghai, Jiangsu, and Tianjin with UEF varying from 1000 to 2000 km2. On the other hand, coastal provinces such as Fujian, Shandong, Liaoning, and Hebei recorded UEF areas below 500 km2.

The area of future UEF scaled up with incremental flood return period (Table S4). Under SSP2-RCP4.5, the BAU scenario indicated that the UEF along coastal China would grow from 5696 (5199–6056) km2 to 9957 (8857–10,542) km2 for 2-year to 50-year coastal flood (the numbers in brackets were the 5%–95% bounds on the SLR projections; same hereafter). The number would increase to 10,940 (10,243–11,549) km2–13,424 (12,997–13,981) km2 for 100-year to 1000-year coastal flood. In comparison to 2020, the 1000-year flood event would see an additional area of 3545 (3118–4102) km2 affected. The average annual change rate during 2020–2050 for this event was estimated to be 1.03% (0.92%–1.16%), which was 1.49 (1.33–1.68) times that of the non-exposed urban land (0.69%) in coastal China. Under SSP5-RCP8.5, the 2-year to 50-year flood would inundate 6587 (6100–7072) km2 –11,383 (10,091–12,217) km2 urban land, and there would be 12,467 (11,762–13,191) km2 –15,002 (14,504–15,616) km2 of urban land affected in the event of 100-year to 1000-year flood. For 1000-year flood, the average annual change rate of UEF was 1.40% (1.29%–1.54%), which was 1.32 (1.21–1.45) times that of the non-exposed urban land (1.06%).

The distribution of UEF varied among provinces under the BAU scenario over the next 30 years (Table S2, Fig. 5). Specifically, in the face of a 1000-year flood event under the SSP2-RCP4.5, Zhejiang showed the largest increase in UEF, with an area of 927 (866–998) km2, followed by Shanghai (704–921 km2) and Guangdong (669–1001 km2). Most of the remaining provinces experienced less than 500 km2 of UEF increment. Under SSP5-RCP8.5, Zhejiang (1224–1378 km2 added) and Shanghai (984–1213 km2 added) still exhibited the most substantial growth in UEF when 1000-year flood occurred. Guangdong (889–1239 km2added) and Jiangsu (776–885 km2 added) followed closely behind, while the rest of the provinces recorded additional UEF mostly below 500 km2.

Fig. 5
figure 5

Increased area of exposed urban lands for 1000-year flood under BAU during 2020–2050 (median estimates with 5%–95% bounds)

Full size image

The UEF under Sustainable Development planning during 2020–2050

Under SSP2-RCP4.5, in the event of 2-year to 50-year flood, 5355 (4887–5711) km2–9512 (8445–10,075) km2 of urban land would be submerged in the coastal China under SD planning, and 10,445 (9779–11,028) km2–12,822 (12,415–13,340) km2 for 100-year to 1000-year flood (Table S4). Specifically, in comparison to the UEF area in 2020 (9879 km2), the 1000-year flood event under SD planning would witness an additional 2943 (2536–3461) km2, with an average annual change rate (0.76%–1.01%) 1.38–1.83 times higher than that of the non-exposed urban land (0.55%) within the study area. Under SSP5-RCP8.5, the area of UEF would be projected to range from 6105 (5625–6578) km2 to 10,840 (9539–11,677) km2 when facing 2-year to 50-year flood, this would be 11,934 (11,235–12,640) km2 –14,414 (13,927–14,987) km2 in case of 100-year to 1000-year. For the 1000-year flood event, the average annual change rate of UEF over the next 30 years was estimated to be 1.27% (1.15%–1.40%), which was 1.49 (1.35–1.65) times that of the non-exposed urban land (0.85%) within the study area.

Taking the 1000-year flood as an example, under SSP2-RCP4.5, Zhejiang would observe the largest increment in UEF, from 2167 km2 in 2020 to 2952 (2899–3022) km2 in 2050, totaling an increase of 785 (732–855) km2, which accounted up to 26.67% (24.70%–28.86%) of the total newly added UEF (Table S3, Fig. 6).

Fig. 6
figure 6

Increased area of exposed urban land for 1000-year flood under Sustainable Development during 2020–2050 (median estimates with 5%–95% bounds)

Full size image

This was followed by Shanghai (624–826 km2 added) and Guangdong (471–765 km2 added), while the remaining provinces saw an increase in UEF less than 500 km2. Under SSP5-RCP8.5, the newly added UEF in Zhejiang (1313–1468 km2 added) and Shanghai (979–1203 km2 added) were roughly above 1000 km2, followed by Guangdong and Jiangsu, with an additional area ranging from 500–1000 km2, and the rest of the provinces gained less than 500 km2.

Comparison of future UEF under different scenarios

Flood-risk urban planning had demonstrated the potential to effectively reduce the increment of UEF when compared to BAU scenario and had showed stability in performance across various flood return period scenarios (Fig. 7, Table S4). Under SSP2-RCP4.5, in comparison to BAU scenario, three planning approaches, namely WP, EP, and DD, exhibited the ability to decrease UEF by 266 (253–269) km2–484 (463–515) km2, 24 (21–26) km2 –64 (57–65) km2 and 149 (134–150) km2–253 (233–267) km2 respectively, for 2-year to 1000-year flood. The SD planning effectively mitigated the newly added UEF by 341 (312–345) km2 –602 (582–641) km2 compared to BAU scenario, a reduction of 16.98% (15.63%–18.67%), which was the optimal planning to lower the newly added UEF, while WP planning was the single planning with the notable performance apart from SD planning. Under SSP5-RCP8.5, the WP, DD, and SD planning also showcased their effectiveness in mitigating UEF compared to the BAU scenario, with reduction area of 221 (215–228) km2 –232 (225–261) km2, 188 (176–190) km2–313 (311–314) km2 and 482 (475–494) km2–588 (577–629) km2, respectively. Among which, SD planning consistently outperformed. Yet, when faced with higher urban land demands, EP planning rather augmented the UEF by 180 (140–192) km2–304 (304–307) km2 relative to BAU scenario.

Fig. 7
figure 7

Reduction in flood exposure compared to BAU under SSP2-RCP4.5 (a) and SSP5-RCP8.5 (b), and exposed urban land for 1000-year flood in 2050 (c) (median estimates with 5%–95% bounds)

Full size image

The effectiveness of plannings in minimizing UEF exhibited a spatial disparity, with SD planning demonstrating optimal performance across the coastal provinces of China (Figs. 8, 9). In the case of the 1000-year flood, for instance, WP planning could effectively mitigate additional UEF compared to BAU scenario in most provinces. It achieved the greatest reduction in UEF in Guangdong, where UEF was reduced by 155–205 km2 over BAU scenario. However, WP planning failed in Zhejiang under high development demands, resulting in an increase of 236–239 km2 over BAU scenario. The EP planning displayed limited effectiveness under SSP5-RCP8.5 in most provinces, especially in Shanghai, Zhejiang, and Guangdong, where experienced an incremental in UEF of 167–180 km2, 107–112 km2 and 45–50 km2, respectively, compared to the BAU scenario. The DD and SD plannings successfully minimized UEF in most provinces, in particular, SD planning fared best in Guangdong, where UEF was lowered by 189–327 km2 compared to BAU scenario.

Fig. 8
figure 8

Exposed urban land for 1000-year flood (50%) under SSP5-RCP8.5

Full size image
Fig. 9
figure 9

Urban exposure for 1000-year flood under BAU (a) and its variation compared to BAU under flood-risk based urban planning (b–e) in each province (median estimates with 5%–95% bounds)

Full size image

Discussion

Relative contributions of urban growth and climate change to the increasing flood exposure

While the changes in UEF in coastal China were driven by a combination of flood variation (due to climate change) and urban growth, the relative contribution of urban growth was generally greater than that of flood variation, which was particularly evident under SSP5-RCP8.5 (Fig. 10). Faced with 2-year to 1000-year flood, if no intervention was taken (BAU scenario), 49.12%–77.07% of newly added UEF was caused by urban growth under SSP2-RCP4.5, and the proportion would grow to 57.37%–80.90% under SSP5-RCP8.5. In case of SD planning, urban growth accounted for 43.24%–72.65% of the rise in UEF under SSP2-RCP4.5, and 52.52%–78.54% under SSP5-RCP8.5. These proportions were lower than those of the BAU scenario to a certain extent.

Fig. 10
figure 10

Relative contributions of urban growth (under BAU and Sustainable Development) and flood variation (with 50% sea level rise scenario) to urban flood exposure in 2050

Full size image

Rapid urbanization spurred dramatic land use changes and the subsequently scaled-up buildings and infrastructures (Ma et al. 2014), which led to urban growth as the dominant driver of changes in UEF in coastal China. China had formed a coastal-oriented development pattern since the reform and opening-up (Liu et al. 2013), where large-scale reclamation projects were carried out to alleviate the pressure of land scarcity (Wang et al. 2020, 2021). Unfortunately, these projects have often encroached upon vast wetlands (Sun et al. 2015), resulting in a disproportionate distribution of urban land in the coastal floodplains (Du et al. 2018). Our findings supported the notion that urban land in coastal China had been and would continue to augment at a brisk rate. This growth served as the primary influencing factor causing future changes in UEF, potentially exacerbating flood exposure and flood risk (Sajjad et al. 2018). Hence, it is imperative to prioritize sustainable urban planning that promotes social disaster preparedness and mitigation as well as the enhancement of urban resilience to flood disasters accordingly.

Policy implications

Efficient utilization of urban land should be noted from a perspective of flood hazard reduction in the coastal China. An unorganized urban sprawl may directly pose the new urban lands to flood-prone areas thus increasing flood exposure and risk (Thieken et al. 2016; Willner et al. 2018). There is widespread concern about urban sprawl globally (Liu et al. 2020a), and compact development aims to utilize land efficiently by controlling the amount of newly added urban land (Mustafa et al. 2018b; Chakraborty et al. 2022). Accordingly, we designed the densification development planning by cutting down on the demand of additional urban land in the future, which was observed to excel in decreasing newly added UEF compared to taking no intervention (Fig. 7), and showed a relatively stable effectiveness across coastal provinces in China (Fig. 9). Similar findings were observed in a study on future flood exposure assessment in Southeast Asian countries, including Thailand, Myanmar, and Vietnam, where urban intensification was identified as an effective strategy for mitigating the impact of urban development on flooding (Tierolf et al. 2021). Additionally, this strategy may occupy a key position in preventing larger fragmentation of forests, farmlands and key wildlife habitats, providing more social wellbeing (Van Berkel et al. 2019). Therefore, limiting urban sprawl in coastal China is essential for mitigating urban flood exposure and establishing sustainable development (Broitman and Koomen 2015, Kaur et al. 2020).

Wetlands protection has been proven effective in curbing coastal UEF. The WP planning that we devised demonstrated the ability to decrease UEF across multiple flood return periods (Fig. 7), reducing newly added UEF by 8.61%–14.93% compared to the BAU scenario. Yet, a substantial proportion of wetlands within China's coastal flood-prone areas had been encroached by urban development over the past two decades, which would further magnify flood exposure and threaten coastal ecosystems (Wu et al. 2022). Currently, the Chinese authorities have been prioritizing the conservation and restoration projects of coastal zones, emphasizing the implementation of coastal eco-construction to strengthen the wetland resistance against flooding hazards. As such, relevant stakeholders should maximize the retention of coastal wetlands in formulating development plans (Choi et al. 2018) to minimize the influence caused by floods (Wang et al. 2014). It is worth noting that there exists a spatial discrepancy in the effectiveness of wetlands protection policies. The WP planning yielded positive results under both SSP2-RCP4.5 and SSP5-RCP8.5 at the national scale, whereas it failed in Zhejiang with rapid development (Fig. 9). Localized policies should be implemented in conjunction with scientific insights. Xu et al. 2021b

Ecological protection strategy necessitates meticulous consideration of the tradeoffs and synergies with flood risk management. In January 2020, the Ministry of Natural Resources of China issued the Guidelines on Territorial Spatial Planning (Trial), which established a series of basic principles and requirements for territorial spatial planning, such as “ecological priority and green development” (He et al. 2022a). However, from a flood alleviation perspective the green-only strategy might not be universally applicable. Although the EP planning aimed to safeguard vegetation, it resulted in a slight increase in UEF relative to BAU scenario under SSP5-RCP8.5, especially in Shanghai, Zhejiang, and Guangdong (Fig. 9). This was because most of the vegetation constraints in EP planning that were distributed in relatively high-elevation regions (200.88 m in average). To meet the escalating urban demand under high-speed development, more urban lands would like to encroach upon the periphery of vegetation reserves in lower-elevation regions (108.97 m in average), which were more susceptible to flood, resulting in an increase in UEF. Similarly, Sun et al. (2022) noted a trend for urban expansion towards flood-prone areas when ecological environment protection constraints were imposed in Shanghai, implied a similarity between our findings, thus corroborating the reliability of our results. Therefore, in formulating relevant policies, more attention should be paid in identifying the optimal and sustainable urban planning strategies that address both ecosystem protection and flood exposure mitigation.

Sustainable urban planning requires comprehensive and systematic approaches to address the complex dynamics of flood exposure (IPCC 2022). Our findings indicate that WP, EP, and DD planning generally perform well and exhibit a synergistic effect, that’s, SD planning shows more consistent effectiveness in controlling UEF increase under both of the SSP2-RCP4.5 and the SSP5-RCP8.5 (Figs. 7, 9). Du et al. (2020) also demonstrated in Shanghai that a “hybrid” strategy outperformed both hard and soft strategies in terms of minimizing flood risk and maximizing benefit/cost ratios. These findings underscore the importance of employing the combination of adaptation measures in urban flood risk management, emphasizing the need for a systematic approach to flood prevention and preparation. Hence, it is crucial to develop and implement integrated flood adaptation measures and spatial planning policies in future urban development.

Limitations and future perspectives

We proposed an approach for promoting sustainable urban development in coastal China amidst climate change. This approach combined the spatial constraints of wetland preservation and vegetation protection, as well as considered the quantity limitations by implementing densification development into future urban planning. This integrated approach aimed to lower newly added UEF compared with no interventions. The sustainable development planning was simultaneously applicable to urban growth scenarios under different socio-economic development pathways and multiple flood scenarios under climate change contexts, thereby, contributing to the enhancement of regional flood adaptive capacities.

Nevertheless, this study has several limitations. First, future flood inundation is affected by a variety of factors such as natural precipitation, topography, ground subsidence, and flood protection level (Du et al. 2020), and we only considered the extent of coastal flood inundation in China under the influence of tides, storm surges, extreme sea level, and topography, which introduces a level of uncertainty. Second, urban growth is driven by multiple inherent factors (Seto et al. 2012; Li et al. 2019; Chen et al. 2020a), and we solely took population as a single driving factor, without giving consideration to the potential impacts of climate change, socioeconomic factors, etc. on urban growth. Even so, our results suggest that UEF in coastal China will continue to mount rapidly. Third, we simply set up the EP planning based on NDVI index without considering other indicators such as permanent basic farmland (Sun et al. 2022), Net Primary Productivity (Liu et al. 2020b), Intact Forest Landscapes (Potapov et al. 2008), and Key Biodiversity Areas (Yang et al. 2020). Moreover, we did not account for the impacts of flood exposure and flood regulation changes on flood risk due to urban growth (Wu et al. 2019). It is important to note that the simulation outcomes do not perfectly represent the actual urban growth and flood exposure in the future, but still hold certain scientific and reference value in terms of reflecting the divergence in urban land patterns and flood exposure control under various planning strategies.

In future studies, it is recommended to apply higher-resolution data and incorporate more driving factors to improve the accuracy of the flood inundation extent and the simulation precision of urban growth within the study area (Zhuang et al. 2022). Meanwhile, employing multiple indicators to design and optimize sustainable urban development plans can enhance the validity and reliability of reducing additional UEF (Shan et al. 2021). Furthermore, we intend to combine SCS-CN (Soil Conservation Service Curve Number), InVEST (Integrated Valuation of Ecosystem Services and Trade-offs) to synthesize and analyze the relationship between urban growth, urban flood exposure and the supply and demand of flood regulation services in the context of climate change (Xu et al. 2020; Zhao et al. 2023b), so as to evaluate urban flood risk and its impacts under different scenarios comprehensively.

Conclusion

Coastal flood exposure and risk along with climate change and urbanization are key elements affecting urban landscape sustainability. This study integrated an urban expansion model, scenario analysis, and flood exposure assessment to quantify the divergence and effectiveness of five urban planning strategies in mitigating UEFs under different socio-economic and climate change scenarios in coastal China during 2020–2050. Even with a moderate scenario of SSP2-RCP4.5, the 1000-year flood UEFs along coastal China was expected to grow by 35.88% (31.56%–41.52%) from 9,879 km2 in 2020 to 13,424 (12,997–13,981) km2 in 2050. Compared to the BAU scenario, each planning strategy was able to prevent additional UEFs in most cases, but it worked best if the spatial and quantitative constraints (Wetlands Protection, Ecological Protection, and Densification Development) were jointly considered for a sustainable development strategy. By applying the Sustainable Development planning strategy, the projected UEFs in 2050 would be reduced by 16.98% (15.63%–18.67%) in a 1000-year flood scenario.

Therefore, the ways of urban growth matters in terms of affecting food exposure and risk, particularly when urban growth was identified as the leading driver of changes in UFE in coastal China. Sustainable ways in other perspectives may increase flood exposure; for instance, urbanization extended into flood-prone areas under ecological protection planning. Synergy should therefore be strengthened across different goals of landscape sustainability. An incorporation of flood exposure control into urban planning should be a foundation to a sustainable landscape. We believe the methodology and the findings could help to facilitate a more comprehensive understanding of sustainable urban landscape.