1 Introduction

In contrast to the gradual formation of dissolution and cover-subsidence sinkholes, abrupt cover-collapse sinkholes present a considerable challenge for both scientists and managers (Han & Hwang2017). These phenomena can cause significant harm to human safety, the natural environment, and urban infrastructure (Tufano et al. 2022). Research on these sinkholes primarily aims to enhance understanding of the development and expansion of subterranean cavities (Sato et al. 2019), with soil composition and groundwater presence identified as critical factors (Ali & Choi 2019). For example, karstic rocks may dissolve in the presence of CO2-rich groundwater, and collapsing Loess formations are often associated with groundwater (Gutiérrez et al. 2019; Jin et al. 2021). Additionally, numerous cases have shown that sinkholes can result from the depletion of confined aquifers, leading to the compaction of loose sand strata over time (Youssef et al. 2020).

In urban settings, shallow sandy layer often causes cover-collapse sinkholes due to its susceptibility to liquefaction and disturbance (Sato & Kuwano 2013). It's disrupted by water ingress from leaking pipelines and surface water accumulation (Ali & Choi 2019). Groundwater extraction, usage of subterranean spaces and intense rainfall are considered dynamic triggers for these sinkholes due to altering groundwater hydraulic gradients (Sato & Kuwano 2013; Lee et al. 2016; Sato et al. 2019; Kwak et al. 2020; Kuwano 2021). Aging underground pipelines are a global concern (Kang et al. 2017). Urban cover-collapse sinkholes caused by these factors have been reported globally, including Jeju, South Korea (Jaya 2015), southwest Japan (Julian 2016), Kuala Lumpur and Ampang Jaya (Rosdi et al. 2017), San Antonio, Texas (Steve & Haworth 2016), Oakwood, Georgia, and Tracer, Colorado in the USA (Roger 2017).

Our study conducted an in-depth analysis of cover-collapse sinkholes in Shanghai over the past two decades, examining their spatial and statistical characteristics in relation to the shallow sand layer, ground elevation, and surface water coverage across the city. The aim was to identify key factors influencing the formation of sinkholes in Shanghai and to recommend approaches for their management and further investigation.

2 Geological and urban constructive setting

Situated at the Yangtze River estuary, Shanghai rests atop unconsolidated sediments with depths ranging from 200 to 500 m. The metropolis has experienced significant urbanization; by 2015, constructed areas represented 46% of its total land, approximately 6833 square kilometers. Within the central urban zone of 664 square kilometers, more than 95% of the land has been developed. Due to their vintage, the majority of subterranean pipelines in this central zone were installed from the 1980s to the early 2000s, which has rendered them increasingly outdated (Zhu et al. 2018). Over the last two decades, Shanghai has witnessed more than 100 ground collapse events (Cai 2019), with an average of more than four occurrences annually. These incidents have substantially undermined both road safety and resident security, necessitating intervention by municipal authorities.

Recent studies (e.g., Zhu et al. 2018; Cai 2019; Cai et al. 2021) suggest that subsurface sinkholes in Shanghai may be triggered by the liquefaction of the shallow sand layer due to underground pipeline leaks. However, the lack of comprehensive analysis concerning soil properties and environmental conditions at these sites means that the exact mechanisms of sinkhole formation in Shanghai are still unclear. Further research is essential to improve safety management of urban infrastructure and to contribute valuable knowledge to the global study of urban cover-collapse sinkholes.

3 Data sources and methods

3.1 Data sources

We utilized Cai's (2019) research as a foundation to construct a comprehensive map documenting sinkhole incidents in Shanghai from 1999 to 2019. Additionally, we integrated data from recent news reports spanning 2020 to 2022 into our database to ensure a complete representation of the city's sinkhole activity. Despite challenges in identifying some data points due to the poor quality of images in the literature, we successfully digitized the locations of 90 sinkholes, including six newly reported cases (see Fig. 1a). Our study also compiled over 3,000 geological boreholes covering the entire Shanghai region (see Fig. 1b). By examining the characteristics of geological stratification, we produced a contour map that illustrates the distribution of shallow sand layer thickness across Shanghai's extensive area of 6,833 square kilometers, achieving an average density of one borehole per two square kilometers.

Fig. 1
figure 1

a Shanghai's location and distribution of cover-collapse sinkholes. Remote sensing image sourced from the Geospatial Data Cloud Platform, China Academy of Sciences, 2017. b The distribution of more than 3000 geological boreholes throughout the city of Shanghai utilized to ascertain the thickness distribution of the shallow sand layer

In 2017, we obtained precise ground elevation measurements using the leveling method, with the Wusong datum as the reference elevation plane. Data points, placed approximately 20 m apart, achieved a precision of ± 1 m and were used to develop high-resolution digital elevation models (DEMs). Additionally, in 2020, we acquired spatial configurations of river surface water through the Third Survey of Land and Resources of Shanghai, employing advanced remote sensing technology with a resolution of 0.5 m. Geological borehole data, ground elevations, and riverine water surface data were sourced from the Shanghai Institute of Geological Survey.

3.2 Methods

For data processing, a MapGIS-based information platform and software ArcGIS are employed. The MapGIS-based information platform, deployed in the Shanghai Institute of Geolgogical Survey, is used to standardize the strata of all acquired boreholes in this research, and then to construct the thickness distribution map of the shallow sand layer. ArcGIS facilitates the digitization of sinkhole locations and conducts spatial analyses to overlay datasets concerning the thickness of shallow sand layers, ground elevation, and riverine surface water coverage. Simultaneously, the density of riverine surface water is calculated by determining the proportion of water area within each 500 m × 500 m grid, drawing on the distribution of city rivers and lakes, with an accuracy of 0.01%. Additionally, Microsoft Excel is utilized to analyze correlations between sinkhole occurrences and variables such as shallow sand layer thickness, ground elevation, and adjacent water areas, and to produce histograms that illustrate these relationships.

4 Results

Over the last twenty years, approximately 87% of sinkhole events have been concentrated in Shanghai's central urban region (Cai 2019, Fig. 1a). In contrast, only a few isolated cases have been recorded in peripheral areas (Fig. 1a). To investigate potential contributing factors like surface water, ground elevation, and the presence of shallow sand layers, we conducted comprehensive spatial and statistical analyses based on collected sinkhole data (Figs. 23, 4).

Fig. 2
figure 2

Spatial distribution of shallow sand thickness and sinkholes in Shanghai (left) and their correlation (right)

Fig. 3
figure 3

Spatial distribution of surface elevation and sinkholes in Shanghai (left) and their correlation (right)

Fig. 4
figure 4

Spatial distribution of river water density and sinkholes in Shanghai (left) and their correlation (right). River water density represents the proportion of river water area within a 500 m × 500 m grid

4.1 Distribution of the shallow sand layer

Shallow sand layers are primarily distributed along the Yangtze River and Hangzhou Bay, with sparse occurrences in central and western Shanghai. The thickness of these layers varies significantly across the region, reaching over 15 m on Chongming Island in northern Shanghai. In contrast, the southeastern part of the city has sand layers between 7 and 15 m thick. Meanwhile, the southern bank of the Yangtze River, the northern bank of Hangzhou Bay, and some urban inland areas feature comparatively thinner layers (see Fig. 2).

Sinkhole occurrences in Shanghai's northern and southeastern districts, characterized by extensive shallow sand layers, are notably sparse. Notably, data reveal an inverse relationship between the prevalence of sinkholes and the depth of these sand layers: of the 90 identified sinkhole sites in the city, 46% are located in regions devoid of any shallow sand layer. The incidence of sinkholes decreases when the sand layer exceeds 3 m in thickness and diminishes significantly at depths beyond 11 m. It is also significant that regions with a sand layer thickness of less than 3 m seldom experience sinkholes (see Fig. 2).

4.2 Distribution of the ground elevation

The geomorphology of Shanghai is predominantly composed of lowlands, generally ranging from 2 to 5 m in altitude. Altitudinal variations are evident across the city, with the western and southwestern regions being the lowest at approximately 3 m, while the central urban area is slightly higher at 4–5 m. The central urban area, especially west of the Huangpu River, is mainly characterized by low ground, with many historical sinkholes and ground elevations around 3 m. In contrast, the three northern islands, including Chongming Island, have elevations primarily between 3 and 4 m. Among the 90 analyzed historical sinkholes, 48 (53%) were observed at a ground elevation of 4 m, while 27 (30%) were documented at 3 m. Notably, 11 sinkholes (12%) occurred at 5 m, three sinkholes at 2 m, and one sinkhole at 6 m.

4.3 Distribution of the riverine surface water

The surface water system in Shanghai is complex, encompassing over 600 square kilometers of surface water within its land boundaries, which accounts for approximately 10% of the total land area. Rivers are spaced at intervals of 100–300 m on average. We delineate the proportion of river water area within a 500 m × 500 m grid as river water concentration. Examination of river water concentration distribution indicates that the highest densities are found in the low-lying western regions, the southern coastal areas, and around Chongming Island in the north, with most areas exceeding 20%. In contrast, the northwestern and eastern parts of Shanghai and central Chongming Island have densities ranging from 8–20%. The central urban area and its surroundings exhibit the lowest density, generally below 4% (Fig. 4).

Out of 90 sinkholes, 51 (57%) are not adjacent to any rivers, while the remaining sinkholes are dispersed across regions with varying river water densities. More sinkholes are found in areas with lower riverine water densities: 10 in areas with 0–2% density, 6 in areas with 2–4% density, and 9 in areas with 4–6% density. No more than five sinkholes are present in areas where river water density exceeds 6%. Overall, the number of sinkholes decreases as river water density increases (Fig. 4).

5 Discussion

5.1 Impact of shallow sand layer

The shallow sand layer in Shanghai, formed during the late Holocene at the confluence of the Yangtze River and the East China Sea (Wei et al. 2010), is distributed along the modern estuary shoreline (Fig. 2). Contrary to previous findings suggesting a link between shallow sand layers and sinkhole formation (Cai 2019), no significant positive correlation exists. Despite the absence of shallow sand layers in many central urban areas, sinkholes remain common.

Our findings indicate that the shallow sand layer may not be as crucial to sinkhole formation in Shanghai as previously thought (e.g., Cai 2019). Recent research (Soliman et al. 2018; Pan et al. 2018; Talib et al. 2022) suggests that confined aquifers surrounded by cohesive clay are more significant in sinkhole occurrence. Cohesive clay resists liquefaction and erosion, indicating that factors like subsurface pipeline leakage and underground space utilization may have a greater impact on sinkholes in Shanghai. However, our understanding of sinkhole formation is limited by the resolution of the current shallow sand distribution map, which has one borehole per 2 square kilometers, while sinkholes are typically smaller than 20 square meters (Cai 2019). The Yangtze River estuary has experienced complex geological changes over time, resulting in diverse spatial distribution of shallow strata, which the existing map cannot accurately represent. Nonetheless, to further ascertain the exact contribution of the shallow sand layer, an in-depth investigation of the geological conditions, particularly the subsurface soil profile (Ali, 2019), pore water pressure (Ishibashi et al. 1977; Mohammadi & Qadimi 2015; Yang et al. 2022), groundwater change (Jia et al. 2018) and other engineering geological indicators (Xia et al. 2019), within the impact zone of the sinkhole, is imperative.

5.2 Impact of ground elevation

Surface water accumulation from precipitation can alter the shallow groundwater environment, increasing the likelihood of sinkhole formation (Delle 2022), particularly in low-lying areas. Our research shows that the central urban area of Shanghai, where sinkholes are common, has a predominantly low terrain, with ground elevations around 3 m, indicating high susceptibility to water accumulation. However, while 30 sinkholes are found at elevations of 3 m, a significantly larger number (60) occur at elevations of 4 m or higher (Fig. 3). This suggests that surface water accumulation from low-lying terrain may have a regional rather than site-specific impact. Additionally, sinkholes at specific sites are largely influenced by subsurface engineering structures that enhance underground hydraulic gradients (Sato & Kuwano 2013). Moreover, the frequency of sinkholes follows a parabolic distribution with ground elevation. As elevation exceeds 4 m, the number of sinkholes decreases (Fig. 3), supporting the idea that higher terrain is less prone to water accumulation.

5.3 Impact of riverine water

Shanghai's estuarine region, influenced by the Yangtze River and the East China Sea (see Fig. 1a), has shallow soils susceptible to river water infiltration. This process, as noted by Lu et al. (2022), can lead to the formation of underground cavities. The presence of surface water near sinkholes indicates the associated risk level. The calculated river water density within each 500 m × 500 m grid (see Fig. 4) reveals that the central urban area, where sinkholes are most common, has the lowest river water density, sometimes reaching zero. More than half of the city's sinkholes lack nearby riverine water. In regions with riverine water, there is a negative correlation between water density and sinkhole occurrence. Larger riverine areas correspond to fewer sinkholes (see Fig. 4). These findings suggest that dense river networks may mitigate shallow soil disturbances by facilitating drainage and indirectly diminish the aging pipeline-caused sinkholes, which frequently occur in urban regions globally (Ali & Choi 2021). Thus, promoting river dredging is an effective strategy to prevent sinkholes in Shanghai, and may be beneficial for other coastal cities globally.

6 Conclusions

This investigation scrutinized data pertaining to sinkholes documented during the past two decades in Shanghai, establishing correlations between these occurrences and the city’s shallow sand layer, ground elevation, and proximity to surface water. The findings suggest that the shallow sand layer plays a modest role, whilst the absence of adjacent rivers, obstructed by urban expansion, plays a significant role in the formation of cover-collapse sinkholes in Shanghai. It is suggested that measures be implemented to enhance river administration, conduct high-precision geolological survey and monitoring, and reinforce research on soil erosive patterns resulting from pipeline leaks and subsurface construction activities.

(1) Approximately 87% of sinkholes in Shanghai over the past two decades are located in the central urban area, characterized by low terrain, thin shallow sand layers, and a scarcity of rivers. This highly urbanized region has been extensively developed, with most of the oldest pipelines embedded underground. It is hypothesized that the formation of cover-collapse sinkholes in Shanghai is primarily influenced by the leakage of aging underground pipelines and engineering activities.

(2) At the site level, most sinkhole locations in Shanghai exhibit three primary characteristics: (a) the shallow sand layer is either absent or ranges from 3 to 11 m in thickness; (b) the ground elevation is typically low to moderate, being 4 m or less; and (c) the density of surrounding river surface water is extremely low.

(3) We propose that cover-collapse sinkholes in Shanghai are primarily instigated by aging pipeline leakage and underground engineering. These factors may disrupt groundwater circulation, leading to sinkholes emerging in unforeseen locations with higher elevations and no shallow sand layers. Most sinkholes are not caused by shallow sandy liquefaction but by cavity formation in the lower confined aquifer, which is overlain by cohesive soil. The absence of natural river channels may impede surface water diversion, contributing to surface water accumulation and indirectly influencing cover-collapse sinkhole formation.

(4) To mitigate the adverse consequences of cover-collapse sinkholes in Shanghai, it is urgently recommended to enhance river dredging and drainage systems to reduce surface water accumulation. Additionally, a high-resolution geolgical survey of subsurface and deep sediments in the central urban area is imperative. Monitoring and research on soil erosion patterns caused by pipeline leaks and underground construction activities must also be strengthened. Such strategies could serve as references for other coastal cities in the world.