Revealing the lagged relationship between groundwater changes and land subsidence in Tianjin based on GRACE and InSAR observations Authors: Yang Li, Wenmiao Zhao, Kuanzhen Mao, Wentao Jing, Weiqi Huo, Dongming Zhang Published: 2026 ABSTRACT AbstractTianjin has long been plagued by land subsidence, which is primarily caused by groundwater over-exploitation and poses major threats to urban construction, economic development, and infrastructure safety. This study aims to explore the relationship between groundwater storage (GWS) changes and land subsidence in Tianjin by combining GRACE and InSAR technologies. To clarify this relationship macroscopically, GRACE and GLDAS data were used to analyze GWS changes in the study area from January 2023 to February 2025, while PS-InSAR was applied to obtain concurrent land subsidence information. Time-lagged cross-correlation analysis was introduced to quantify the correlation and lag time between GWS changes and land subsidence time series. Results showed: (1) GWS was positive in most areas (maximum annual rate 59.13 mm/yr, concentrated in east-central Jinghai, south-central Xiqing, Jinnan and central Binhai New Area), decreasing from the center to north and south, with a significant decline (-9.22 mm/yr) in northern Binhai New Area; severe land subsidence showed a “peripheral aggregation” pattern, with 4 contiguous zones, 7 funnels and a max rate of 67.62 mm/yr (near Yangjiapo Town, Binhai New Area); (2) The cross-correlation coefficient between GWS changes and land subsidence ranged from 0.257 to 0.882 (0.542–0.853, 0.431–0.826, 0.257–0.882, 0.538–0.846 for Zones I-IV respectively), and land subsidence lagged GWS changes by 41–263 days overall. This study confirms the link and lag effect between GWS and land subsidence in Tianjin, providing scientific guidance for subsidence prevention and early warning.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-58032-w.Keywords: GRACE, PS-InSAR, SBAS-InSAR, Groundwater storage, Land subsidence, Cross-correlationSubject terms: Environmental sciences, Natural hazards Introduction Land subsidence is a geological disaster characterized by the decline of surface elevation caused by human activities and natural factors. It is imperceptible, irreversible, wide-ranging in impact, and difficult to prevent and control, potentially leading to irreversible losses of environmental resources and posing severe threats to residents’ lives, economic development, urban construction, and infrastructure1. The North China Plain, where Tianjin is located, is one of the regions with the fastest subsidence rates and the largest impact scope in China2. As an alluvial plain area, Tianjin has experienced large-scale subsidence since the Cenozoic era of the plain; it is among the earliest and most severely affected regions by land subsidence, with the first discovery of land subsidence dating back to 1932. Land subsidence in Tianjin may trigger geological disasters such as subsidence funnels, collapse pits, ground fissures, and landslides, seriously hindering the economic development and construction of the region3,4. Therefore, it is of great significance to conduct continuous land subsidence monitoring in Tianjin. For land subsidence monitoring, traditional methods (e.g., leveling, layer-wise mark measurements, and GNSS) can achieve high temporal resolution and measurement accuracy. However, these methods rely on expensive equipment and only target specific points, making it difficult to obtain large-scale land subsidence monitoring information. Interferometric Synthetic Aperture Radar (InSAR) technology, developed in the 1980s, compensates for the shortcomings of traditional measurements, featuring all-weather operation, wide monitoring coverage, and high monitoring accuracy5. Nevertheless, conventional InSAR technology is affected by errors such as spatio-temporal decorrelation and atmospheric delay, which limit its monitoring precision. To obtain higher-precision long-term time-series deformation information, scholars have successively proposed two main categories of InSAR time-series analysis methods: Small Baseline Subset InSAR (SBAS-InSAR)6,7 and Persistent Scatterer InSAR (PS-InSAR)8–10. The selection of an appropriate InSAR time-series method is primarily determined by the dominant scatterers in SAR images: SBAS-InSAR is suitable for regions where distributed scatterers are prevalent, while PS-InSAR is preferred in areas with dominant persistent scatterers. PS-InSAR extracts highly coherent and stable scatterers, typically stable ground features such as corners of man-made buildings and bridges11. Due to the low sensitivity of these Persistent Scatterer (PS) points to spatio-temporal decorrelation, analyzing the phase composition of these stable points and leveraging the characteristics of each phase component to separate terrain, flat-earth, and atmospheric phases enables the acquisition of accurate surface deformation phases and further calculation of deformation information. This method performs well in urban subsidence monitoring, achieving millimeter-level precision12. Since the study area is mainly located in the built-up region of Tianjin with a large number of PS points, it is particularly suitable for PS-InSAR. Therefore, this study adopts PS-InSAR as the main technique to retrieve land subsidence information. Groundwater is closely related to land subsidence, and groundwater over-exploitation has long been regarded as the primary driving factor of land subsidence in the North China Plain13. Currently, groundwater level monitoring primarily relies on groundwater wells, which have a limited observation range and cannot meet the needs of large-scale groundwater change monitoring. In contrast, the Gravity Recovery and Climate Experiment (GRACE) satellite, jointly developed by NASA and the German Aerospace Center, provides a new approach for inverting dynamic groundwater changes as it is not affected by terrain, climate, or other factors14–16. Existing studies have shown that GRACE can invert large-scale changes in surface mass, which is sufficient to detect variations in terrestrial water storage (TWS) with an average water height of less than 1 cm. By combining TWS data obtained from GRACE with external data, changes in groundwater storage (GWS) can be inverted17. Scholars in relevant studies on groundwater and land subsidence shifted their focus from point-scale to regional-scale analyses. At the point scale, studies have explored the relationship between groundwater levels in wells and land subsidence at monitoring sites18. At larger scales, research has investigated the spatial consistency between regional groundwater contour lines and land subsidence centers19. In recent years, time-series InSAR techniques combined with wavelet and cross-spectral analysis have been widely used to analyze the coherency and phase delay between surface deformation and environmental drivers20–22. Meanwhile, scholars have begun to integrate InSAR and GRACE technologies to better identify the relationship between regional groundwater and land subsidence23–25. However, existing studies still have certain limitations. Owing to the difficulty in obtaining in-situ groundwater level observations, relevant studies are often restricted to areas with monitoring wells. In regions lacking groundwater data, most analyses focus only on climatic factors such as precipitation and temperature. Although GRACE observations provide a promising alternative, few studies have systematically performed time-lagged cross-correlation analysis between GRACE-derived GWS and InSAR-derived land subsidence in Tianjin, and the spatial heterogeneity of lag times remains poorly understood. To fill these gaps, the main contributions of this study are: (1) To obtain high-precision land subsidence information in Tianjin from January 2023 to February 2025 using PS-InSAR, with verification by SBAS-InSAR. (2) To invert concurrent groundwater storage changes using GRACE and Global Land Data Assimilation System (GLDAS), and validate the results with in-situ groundwater level data. (3) To quantify the correlation and time-lag characteristics between GWS changes and land subsidence using time-lagged cross-correlation analysis. The remainder of this paper is organized as follows: “Materials and methods” Section introduces the study area, data sources, and methodology. “Results” Section presents the results of land subsidence, groundwater storage changes, and time-lag analysis. “Discussion” Section discusses the mechanisms, compares with previous studies, and analyzes the uncertainties and limitations. “Conclusions” Section summarizes the main conclusions of this study. This study is expected to provide scientific references for the prevention, control, and early warning of land subsidence in Tianjin. Materials and methods Study area Tianjin lies between latitudes 38°34′N and 40°15′N, and longitudes 116°43′E and 118°04′E. It administers 16 districts, covering a total area of approximately 11,966.45 km² and having a permanent population of 13.64 million. The average altitude of the urban area is 2–5 m. Tianjin has a warm temperate semi-humid monsoon climate. The mean annual temperature is 12.5 °C, with January being the coldest month (extreme minimum temperature reaching − 17.8 °C) and July the warmest (extreme maximum temperature of 40.5 °C). The mean annual precipitation is about 550 mm, and more than 80% of the annual precipitation is concentrated from June to September, which recharges regional groundwater through infiltration. Long-term records show that annual precipitation presents a slightly increasing trend with significant interannual variation26. Tianjin is located in an alluvial and marine plain and is an area with relatively active neotectonic movements. It has complex geological conditions and many geological environment problems, and is one of the cities in the North China Plain and even in China with the earliest and most serious land subsidence13. Land subsidence was first discovered in 1923, and the maximum cumulative subsidence has exceeded 3.4 m so far27. At the same time, Tianjin is also one of the cities in China with relatively scarce water resources. Due to water quality reasons, most surface water cannot be used directly, and the shortage of available surface water resources has restricted the long-term development of the city. Before the implementation of the Luanhe River Water Diversion Project and the Yangtze River Water Diversion Project, the development and utilization of groundwater resources was an important part of the water sources used in Tianjin, and the over-exploitation of groundwater has increased land subsidence in the southern plain area of Tianjin3. According to the “Tianjin Geological Disaster Prevention and Control Plan (2021–2025)”28, “Land subsidence is the most important geological disaster in the plain area of Tianjin. In 2019, the mean subsidence across Tianjin’s plain area reached 19 mm, with the central urban area and Binhai New Area recording 11 mm and 14 mm respectively; the area of serious subsidence areas with an mean subsidence of more than 50 millimeters in the city was 773 km², accounting for 6% of the city’s total area, mainly distributed in the central and northwestern parts of Jinghai District, the junction of the southern part of Wuqing District and the southwestern part of Beichen District, the northeastern part of Hangu in Binhai New Area, and the junction of Xiqing-Jinghai-Dagang”. Therefore, the scope of the study area mainly focuses on the central and southern regions of Tianjin where land subsidence problems are concentrated, including the main urban area (Heping District, Hexi District, Hedong District, Hebei District, Nankai District and Hongqiao District), the peri-urban districts (Beichen District, Dongli District, Xiqing District, Jinnan District and Jinghai District) as well as Binhai New Area, as shown in Fig. 1. Geographic location of the study area. Data source InSAR data The Sentinel-1 A satellite is an environmental monitoring satellite launched by the European Space Agency (ESA) following the ERS and Envi-sat series of satellites. It has 4 imaging modes, among which the interferometric wide swath (IW) mode acquires images with a width of 250 km at a medium resolution of 5 m × 20 m29. In this study, C-band Sentinel-1 A satellite SAR images were used to obtain InSAR time series data in IW mode. Precise Orbit Ephemerides were used to correct the orbit information30. The “ALOS World 3D − 30m” (AW3D30) provided by the Japan Aerospace Exploration Agency (JAXA)31 was used to subtract the topographic phase component and conduct geocoding during time-series InSAR processing32. The generic atmospheric correction online service for InSAR (GACOS)33 was used to correct the atmospheric phase delay during the interferometric processing34. The specific information of the data is shown in Table 1. Table 1The information of InSAR data.NameTemporal rangeResolutionData sourceCountOrbital frame numberPolarizationFlight directionSentinel-1 A2023.01-2025.025*20m https://search.asf.alaska.edu/ 46Path 69, Frame 124VVASCENDINGPOD2023.01-2025.025 cm https://scihub.copernicus.eu/ 46–––AW3D30/30 m https://www.eorc.jaxa.jp/ 1–––GACOS2023.01-2025.0290 m http://www.gacos.net/ 46––– The information of InSAR data. Groundwater data The mascons data from global gravity satellites (GRACE and GRACE-FO) were obtained from the Center for Space Research (CSR)35, Jet Propulsion Laboratory (JPL)36, and Goddard Space Flight Center (GSFC)37. The data processing of mascons includes removing non-surface mass change signals such as atmospheric and oceanic tides, and enhancing signal quality through various filtering techniques. The final product is presented in the form of a global grid, where each grid cell contains time series data reflecting regional mass changes. The GLDAS NOAH land surface model data cover information such as surface runoff, soil moisture, canopy water, and snow water38. In this study, GRACE mascon data from 2023 to 2025 and the corresponding GLDAS-NOAH data were adopted. Both datasets have a spatial resolution of 0.25° and monthly temporal resolution, ensuring full compatibility. To directly verify the reliability of GRACE-derived GWS data for the core study period, in-situ groundwater level data were collected from 5 groundwater monitoring wells in Tianjin, covering the study period with a monthly sampling frequency. This in-situ data provides quantitative validation for the GRACE-inverted GWS changes, ensuring the credibility of subsequent cross-correlation analysis. Methods Research technique route As shown in Fig. 2, the workflow of this study is mainly divided into three parts: InSAR Data Processing, Groundwater Change Data Processing, and Time-lag Analysis of Land Subsidence and Groundwater Change. Workflow of this study. In the InSAR Data Processing part, Sentinel-1 A images, Precise Orbit Ephemerides, AW3D30 DEM, and GACOS data were used as inputs. Deformation information was obtained through PS-InSAR and SBAS-InSAR technologies respectively. The results were compared and verified, and the obtained Line-Of-Sight (LOS) deformation data are adjusted to the vertical direction, which were used as the time-series results of land subsidence. In the Groundwater Change Data Processing module, on the one hand, GRACE data were processed to obtain terrestrial water storage change data; on the other hand, components such as soil water, snow water, and canopy water are extracted and retrieved from GLDAS hydrological model data. GWS changes were estimated based on the water-balance principle, and accuracy verification was carried out by combining with field-monitored groundwater data, finally obtaining groundwater change data. In the Time-lag Analysis part of Land Subsidence and Groundwater Change, after preprocessing such as polynomial fitting was performed on the above-obtained land subsidence data and groundwater change data, time-lagged cross-correlation analysis was used to solve the groundwater change and land subsidence data to obtain the number of days that land subsidence lags behind groundwater change, and the correlation between the two was calculated, providing a scientific basis for the prevention, control and early warning of land subsidence geological hazards in Tianjin. Monitoring land subsidence via InSAR PS-InSAR PS-InSAR The PS-InSAR technology was proposed by Ferretti et al.39. The key of PS-InSAR technology lies in extracting stable target points (i.e., Persistent Scatterers) with high temporal and spatial correlation in the study area using statistical analysis methods when processing multiple SAR images covering the same area acquired at different times. By modeling and analyzing the phase time series of PS targets, deformation information can be retrieved40,41. The specific process is as follows. Suppose N + 1 images are used, one image is selected as the common reference image through calculation, and the others serve as secondary images. All secondary images are registered with the reference image to obtain N interferometric pairs. Differential interferometric processing is performed on these N interferometric pairs, and external DEM data is used to eliminate the influence of topographic factors. From the processing results, the Amplitude Dispersion Index (ADI) value of a single pixel is estimated. For a specific pixel αin a SAR image, its ADI value across the N + 1 SAR images is: where is the mean amplitude of the pixel across all images, and is the standard deviation of the pixel’s amplitude over the time series. PS points in the SAR image are detected by calculating ADI values. Then, the phase information of these reliable PS points is analyzed uniformly. Combined with the above formula, the phase value of PS point in the interferogram can be expressed as: where φ is the interferometric phase of the PS point in the interferogram, which is a wrapped phase; is the deformation phase along the LOS direction; is the topographic residual phase; is the orbital residual phase; is the atmospheric phase; is the noise phase, including spatial or temporal decorrelation noise, system noise, and other noises. PS points are characterized by high stability, and and can be eliminated through filtering and GACOS data. Residual elevation and deformation rate can be estimated based on a specific linear model, while and are removed using external DEM data and satellite precise orbit data. Finally, the deformation phase along the LOS direction is obtained, which is converted to generate the deformation time series of PS points. Ultimately, the one-dimensional deformation information in the LOS direction is decomposed into vertical deformation components, serving as the time-series results representing land subsidence42,43. The processing flow is shown in Fig. 3. PS-InSAR basic processing flow. 2.SBAS-InSAR SBAS-InSAR The SBAS-InSAR technique was proposed by Berardino et al.6. Its principle involves configuring appropriate spatiotemporal baseline thresholds to partition SAR images covering the same area into subsets. Interferometric pairs are generated through interferometric processing. The surface deformation time series is then constructed using the least-squares method. Additionally, the Singular Value Decomposition (SVD) technique is employed to obtain the minimum norm solution, thereby resolving the rank deficiency issue of the small-baseline subset matrix. Meanwhile, the elevation residual phase is estimated, and external atmospheric correction data is used to reduce atmospheric delay errors during processing. This method can form a relatively large number of interferometric pairs even with a small amount of data, enabling the acquisition of high-quality surface deformation information7. It is currently the most widely used method for long-term surface deformation monitoring. The specific process is as follows: Acquire N + 1 SAR images covering the same area, select one of them as the reference image, and then perform registration, resampling, as well as removal of the flat-earth and topographic phases. Based on reasonable spatiotemporal baseline thresholds, M interferograms are generated, where the M interferometric pairs must satisfy the inequality as follow. The phase value of each pixel in the interferogram can also be expressed by Eq. (2). The deformation time series are solved using a least-squares inversion method. When gaps exist in the interferogram network, the SVD technique is applied to estimate a linear trend across the missing connections. The flowchart of this process is shown in Fig. 4. SBAS-InSAR basic processing flow. Obtaining GWS via GRACE and GLDAS Regional terrestrial water storage variability can be effectively obtained by inverting the time-variable gravity fields of GRACE and GRACE-FO through the following equations44,45: where represents the equivalent water height of terrestrial water storage anomalies; A denotes the Earth’s radius (6371.39 km); is the average density of the Earth (5.51 × 103 kg/m3); is the density of water (103 kg/m3); and are the load Love numbers of degree l; W stands for the kernel function of the Gaussian filter; represents the fully normalized associated Legendre functions; and and are the variations in the spherical harmonic coefficients of the Earth’s gravitational field derived from GRACE/GRACE-FO global gravitational models (GGMs). Inversion of GRACE data reveals monthly anomaly changes in terrestrial water storage () relative to the climatological mean of the same calendar month within the study period, which include variations in surface runoff equivalent (), groundwater storage (), soil moisture storage (), snow water equivalent storage () and canopy water storage ()46. Due to the small surface runoff in the North China Plain, the surface water changes across the entire study area are relatively minor, and their impact on the overall inversion results is extremely weak. Therefore, is temporarily excluded from the calculation45,47. Based on the “principle of water balance”, GWS changes can be obtained by subtracting, , and from , through the following equation: where ΔSMS, ΔSWES, and ΔCANS are obtained from the GLDAS_NOAH model. The resolution of GRACE data is 0.25°, and the GWS obtained at this scale reflects large-area anomalies rather than the actual GWS at specific locations. Therefore, the results derived from the processing of GRACE and GLDAS data should be regarded as representing the overall changes in GWS in the region, serving as an indicator of GWS change signals. To verify the reliability of GRACE-derived GWS changes, Pearson correlation analysis was conducted between the inverted GWS time series and in-situ groundwater level observations from the study area. The in-situ groundwater level data were spatially and temporally matched with the GRACE grid scale and monthly temporal resolution of GWS changes. The Pearson correlation coefficient (r) and statistical significance (p-value) were calculated to quantify the linear correlation between the two datasets, where a higher r (close to 1 or -1) and p < 0.05 indicated a significant linear correlation, validating the accuracy of the GWS inversion results. Time-lagged cross-correlation analysis Due to differences in factors such as groundwater depth and aquifer lithology, the response of land subsidence to groundwater changes is usually not immediate. Therefore, it is inappropriate to use correlation coefficients directly for evaluation. In this study, cross-correlation analysis will be used to reveal the relationship between groundwater changes and land subsidence. Cross-correlation analysis is a signal processing method based on black-box or grey-box models, which explores the relationship between the input time series () and the output time series () by introducing a relative shift between them48. The cross-correlation coefficient is defined as follows: where is the cross-covariance of the two series; n denotes the length of the time series; k represents the lag time; is the cross-correlation coefficient; and are the sample standard deviations of the and series, respectively; and and are the sample means of the and series, respectively. The peak value of reflects the sensitivity of the signal response between the two time series. A larger value indicates that is more susceptible to the influence of . Additionally, the lag time corresponding to the first peak of the cross-correlation coefficient represents the time by which one series lags behind the other49. Results Comparison of PS-InSAR and SBAS-InSAR Results Before applying the PS-InSAR deformation results to the analysis of land subsidence in the study area, reliability verification is required. Due to the lack of concurrent ground observation data, this study indirectly evaluated the accuracy of InSAR monitoring results through correlation analysis and consistency assessment of the monitoring results from PS-InSAR and SBAS-InSAR (as shown in Fig. 5). Comparison of PS-InSAR and SBAS-InSAR Results. (a) The result of linear regression analysis. (b) The Bland-Altman consistency plot. The results indicate a high level of consistency between the two methods: in the linear regression analysis,, which demonstrates a strong alignment in the deformation rate monitoring trends of the two methods. The Bland-Altman consistency plot further quantifies the range of differences, with 95% of data point differences concentrated within the interval of -1.19 to 9.78 mm/yr. This difference can be attributed to technical variations in the implementation processes of the two methods, such as discrepancies in automatically derived stable reference benchmarks, differences in connection graph construction strategies and ground control point (GCPs) selection rules, which in turn lead to numerical fluctuations in deformation rates. Additionally, in this study, the highly coherent PS points from PS-InSAR were used to extract pixel values for SBAS-InSAR. However, the dense monitoring point distribution characteristic of SBAS-InSAR may introduce slight deviations in the coherence of PS points within SBAS pixels42. By observing the results of the two methods, the points with poor consistency are mainly distributed in the areas around green spaces, water bodies and roads in the study area; these regions are presumably dominated by distributed scatterers rather than stable persistent scatterers, and the differential adaptability of PS-InSAR and SBAS-InSAR to distributed scatterers may be the key factor leading to the deviation of monitoring results at these locations. In summary, the mutual verification of PS-InSAR and SBAS-InSAR results supports the reliability of InSAR monitoring to a certain extent, and also proves the methodological robustness of InSAR technology in high-precision monitoring of surface deformation from the perspective of technical cross-validation. Land subsidence in Tianjin In this study, the PS-InSAR technique was employed to extract surface deformation characteristic information from Sentinel-1 A satellite data in the study area spanning from January 2023 to February 2025. Furthermore, the LOS deformation information was converted into vertical deformation information, and finally, the land subsidence rates and cumulative subsidence amounts of each subsidence zone were obtained (results shown in Fig. 6). Monitoring results indicated that the severe subsidence zones were mainly distributed around Tianjin’s main urban area, while only mild and localized subsidence existed within the main urban area (-20 to -5 mm/yr). A total of 4 relatively concentrated and contiguous subsidence zones were identified, within which 7 subsidence funnels were developed. Temporal and spatial characteristics of land subsidence in Tianjin. (I) Beichen-Xiqing-Jinghai, (II) Jinnan-Jinghai, (III) Dongli, (IV) Binhai New Area. Subsidence Zone I is primarily distributed in Bazhou City, Hebei Province, the central and western parts of Jinghai District, as well as the western parts of Xiqing District and Beichen District. There are two subsidence funnels in this zone. Funnel I-1, centered on Wangqingtuo Town in Wuqing District and Shengfang Town in Bazhou City, showed a tendency to connect into an approximately elliptical feature tilted in the northeast-southwest direction, with a maximum subsidence rate of 60.05 mm/yr. This area is mainly composed of industrial parks and residential areas, where large-scale land subsidence disasters have occurred historically3,28,50. Funnel I-2 spreads outward from Tuanbo Town as the center, with a maximum subsidence rate of 55.05 mm/yr, and is dominated by residential areas. As can be seen from the subsidence curves, with the passage of time, the cumulative subsidence of Funnel I-1 showed an annual increasing trend, while the cumulative subsidence of Funnel I-2 began to slow down after July 2024. Subsidence Zone II is mainly located at the junction of Jinghai District, eastern Xiqing District, and southern Jinnan District. It contains two subsidence funnels. Funnel II-1 is primarily distributed in Wangwenzhuang Town of Xiqing District, with a maximum subsidence rate of 63.26 mm/yr, and covers areas with structures such as residential quarters and industrial parks. Funnel II-2 is mainly in Balitai Town of Jinnan District, with a maximum subsidence rate of 60.05 mm/yr, surrounded by large areas of farmland and factories. Subsidence curves showed that after May 2024, the cumulative subsidence of both funnels began to slow down. Subsidence Zone III is concentrated in the central part of Dongli District, with two subsidence funnels. Funnel III-1 is mainly in the northern part of Junliangcheng in Dongli District, with a maximum subsidence rate of 45.15 mm/yr. This area has seen extensive construction of new residential quarters in recent years, with surrounding farmland. Funnel III-2 is located in the Airport Economic Zone east of Binhai International Airport, with a maximum subsidence rate of 35.96 mm/yr. This area is a newly built comprehensive park in Dongli District, including numerous industrial parks, commercial office areas, and residential zones. Subsidence curves indicate that the cumulative subsidence of Funnels III-1 and III-2 gradually slows down after January 2024 and July 2024, respectively. Subsidence Zone IV is distributed in the border area between the northern part of Binhai New Area and Tangshan City, with one subsidence funnel (Funnel IV-1) located near Yangjiapo Town, Binhai New Area. The subsidence range spreads outward from the subsidence funnel as the center, with a maximum subsidence rate of 67.62 mm/yr. This area is dominated by a large number of villages and farmlands, with some industrial parks scattered throughout. Subsidence curves showed that the cumulative subsidence of Funnel IV-1 continues to increase, and no slowdown trend has been observed yet. Overall, land subsidence in Tianjin covers a wide range and is spatially uneven. In terms of land use types, it is mainly concentrated in residential areas and industrial parks. Excessive groundwater extraction to meet the demands of domestic water, industrial water, and other uses may be the main reason for the increasing land subsidence in these areas51,52.The maximum subsidence rates in all zones exceeded 35 mm/yr, and there is a tendency for them to connect into a single area and expand outward. It is necessary to take effective measures to control surface subsidence; otherwise, they may develop into more severe subsidence funnels. GWS changes in Tianjin Based on the water balance equation, this study derived the GWS changes in the study area and further generated the GWS change time-series curve. Direct quantitative validation was performed using monthly in-situ groundwater level data (2023–May 2025) from 5 monitoring wells (shown in Fig. 7). Pearson correlation coefficients between GRACE-derived GWS and in-situ groundwater levels ranged from 0.638 (W4, Binhai New Area) to 0.861 (W2, Jinnan District), with all p-values less than 0.05, indicating statistically significant correlations. High consistency (R > 0.8) at wells W2 (Jinnan District) and W3 (Beichen District) indicated that GRACE-derived GWS accurately captures regional groundwater dynamics. The slightly lower correlation at W4 is likely influenced by the coarse spatial resolution of GRACE, which may not fully capture localized groundwater dynamics in the Binhai New Area. This validation strengthens the credibility of the 2023–2025 GWS data for subsequent analyses. Comparison between GWS change and in-situ groundwater level. (a) Location of the groundwater monitoring wells in Tianjin. (b) Temporal comparison between GRACE-derived GWS change and in-situ groundwater level at each well. Comparison results indicated that the two datasets shared consistent overall variation trends and periodic fluctuations, but differ in fluctuation amplitude and the timing of changes. The difference in fluctuation amplitude is attributed to the following: the measured data reflect groundwater changes in local areas (where groundwater exists), and the small area of these regions leads to a relatively high equivalent water height. In contrast, GRACE inversion covers a global scale; according to the principle of mass conservation, large areas without groundwater reduce the average equivalent water height, resulting in a smaller value of the GRACE-derived equivalent water height. The difference in change timing may be explained as follows: the measured data rely on monitoring wells with limited coverage. Even small-scale groundwater extraction or recharge activities in the vicinity of these wells can cause changes in groundwater levels, so the measured data can only explain groundwater changes in specific local areas. However, GRACE data have low spatial resolution, meaning that the inverted results can only capture the temporal changes of groundwater when large-scale water use activities cause significant gravitational anomalies. Notably, the two datasets were quite consistent over a long time span. This indicated that the method of inverting groundwater changes using GRACE data was reliable and feasible for analyzing the long-term variation trends of groundwater. Through the inversion of GRACE data, this study obtained the spatial distribution of the interannual GWS change rate in the study area from January 2023 to February 2025 (as shown in Fig. 8). The results showed that the GWS change rate is positive in most areas of the study area, with a maximum value of 59.13 mm/yr. Specifically, the GWS in the east-central part of Jinghai, the south-central part of Xiqing, Jinnan District, and the central part of Binhai New Area showed a rapid increasing trend. The GWS in the main urban area, Dongli District, and Beichen District showed a slow increasing or basically flat trend. The area with the most significant GWS decline is located near Yangjiapo Town at the border of northern Binhai New Area and Tangshan City, with a change rate as low as - 9.22 mm/yr, which is consistent with the location of the groundwater depression cone marked in the Tianjin Water Resources Bulletin53,54. Annual change rate of GWS. Notably, the positive GWS recovery in east-central Jinghai, south-central Xiqing and Jinnan Districts is likely mainly reflected in the shallow unconfined aquifer, which may be directly replenished by the South-to-North Water Diversion Middle Route Project (SNWDP-MRP) passing through Xiqing District and other key areas55. Taking 2024 as an example, the SNWDP-MRP delivered 9.7818 × 108 m3 of water to Tianjin, with Tianjin’s total groundwater resource volume reaching 9.7013 × 108 m3 in the same year. This groundwater volume was 68% higher than the multi-year average (5.7756 × 108 m3) and 38.9% higher than that in 2023 (6.9842 × 108 m3)53,54, suggesting that the SNWDP-MRP is likely a key driver of the observed GWS recovery in these districts. However, these areas overlap with Subsidence Zone I and II (shown in Fig. 8), where severe subsidence is driven by deep confined aquifer compaction. Previous studies on Tianjin’s subsidence mechanism have confirmed that the contribution rate of deep strata (> 300 m) to land subsidence reaches 65%, with areas like Xiqing and Hangu exceeding 70%13. The thick clay aquitard between shallow and deep aquifers forms a strong hydraulic barrier, preventing effective recharge of deep groundwater by the SNWDP-MRP2,55. This vertical stratification of aquifers and the layered recharge characteristics of the water diversion project lead to the spatial contradiction between the GWS recovery and deep severe subsidence. Time-lagged cross-correlation analysis Excessive groundwater extraction gradually reduces the porosity of aquifers, leading to an increase in effective stress and thus causing land subsidence56. To comprehensively reveal the dynamic response between GWS changes and land subsidence, both the raw time series and the dynamic change indicators were analyzed. Polynomial fitting was conducted on several representative points within each subsidence zone to obtain two sets of time-varying curves: (1) GWS changes and cumulative subsidence (reflecting long-term cumulative trends); (2) GWS rate of change (first-order difference of raw GWS, representing monthly variation intensity) and subsidence acceleration (second-order difference of cumulative subsidence, i.e., rate of change of subsidence velocity, reflecting the acceleration/deceleration of subsidence). The integrated time-series curves are shown in Fig. 9. Temporal variations of GWS, cumulative subsidence, GWS rate of change, and subsidence acceleration in the four subsidence zones of Tianjin (2023–2025). (a) GWS changes and cumulative subsidence. (b) GWS rate of change and subsidence acceleration. As shown in Fig. 9 (a), the fluctuation trends of cumulative subsidence and GWS changes were relatively consistent, with an obvious lag effect between the two. Specifically, when GWS increased (groundwater recharge), the land subsidence rate slowed down accordingly; when GWS decreased (groundwater depletion), the land subsidence rate accelerated. However, due to the lag effect induced by aquifer compaction and pore water pressure dissipation, there was a certain time difference between these two processes. Notably, the more sensitively captures the dynamic lag response. The GWS rate of change consistently exhibited a leading pattern relative to subsidence acceleration across all four zones: the negative peaks (rapid groundwater depletion) or positive peaks (rapid groundwater recharge) of the GWS rate of change appeared in advance, followed by corresponding negative peaks (accelerated subsidence) or positive peaks (decelerated subsidence) of subsidence acceleration. This phenomenon directly confirms that the dynamic change of GWS is the direct driver of the subsequent acceleration of land subsidence, which aligns with the poroelastic response mechanism of aquifer systems. Notably, Fig. 9 (b) more sensitively captures the dynamic lag response. Despite their different physical units, the two curves exhibit highly consistent temporal trends. Land subsidence in the study area is a long-term, persistent deformation process primarily controlled by historical groundwater over-exploitation and lagged creep of thick clayey aquitards; thus, the overall subsidence rate is relatively stable and less sensitive to short-term fluctuations in GWS. In contrast, subsidence acceleration (the rate of change of subsidence rate) effectively amplifies the subtle response of subsidence to short-term GWS variations. The GWS rate of change consistently exhibited a leading pattern relative to subsidence acceleration across all four zones: the negative peaks (rapid groundwater depletion) or positive peaks (rapid groundwater recharge) of the GWS rate of change appeared in advance, followed by corresponding negative peaks (accelerated subsidence) or positive peaks (decelerated subsidence) of subsidence acceleration. This phenomenon directly confirms that dynamic changes in GWS drive the subsequent acceleration or deceleration of land subsidence, which aligns with the poroelastic response and lagged compaction mechanism of the layered aquifer system. Table 2 presents the lag time of land subsidence relative to GWS changes in each subsidence zone, as well as the correlation coefficient between the two. Differences in groundwater level depth across various zones result in variations in the lag time of land subsidence behind GWS changes. Overall, the lag time of land subsidence relative to GWS changes in the study area ranges from approximately 2 to 9 months. A higher correlation coefficient between cumulative subsidence and GWS changes indicates a more significant impact of GWS changes on land subsidence; the relatively low correlation coefficient in some zones is presumably related to factors such as local meteorological conditions and geological structures. Table 2Time-lagged cross-correlation coefficients between GWS changes (mm) and land subsidence (mm) in different subsidence zones.Subsidence zoneLag time (day)Correlation coefficientI97–1860.542–0.853II41–1120.431–0.826III81–2630.257–0.882IV69–1590.538–0.846 Time-lagged cross-correlation coefficients between GWS changes (mm) and land subsidence (mm) in different subsidence zones. Discussion The characteristics of land subsidence in Tianjin The spatio-temporal characteristics of land subsidence in Tianjin from January 2023 to February 2025, derived from PS-InSAR monitoring, reveal distinct heterogeneity, which is closely associated with regional hydrogeological conditions, land-use patterns, and human activity intensity. Severe subsidence is primarily distributed in the periphery of the main urban area, while the main urban area itself experiences more moderate, but widespread, subsidence. A total of 4 contiguous subsidence zones and 7 subsidence funnels have been identified, presenting an obvious “peripheral aggregation” pattern (Fig. 6). Among the four subsidence zones, Subsidence Zone I and Subsidence Zone IV are the most severely affected. Subsidence Zone I contains two large-scale funnels: Funnel I-1, centered on Wangqingtuo Town (Wuqing District) and Shengfang Town (Bazhou City), with a maximum subsidence rate of 60.05 mm/yr; and Funnel I-2, centered on Tuanbo Town (Jinghai District), with a maximum subsidence rate of 55.05 mm/yr. Land-use analysis revealed that Funnel I-1 was dominated by industrial parks (e.g., the bicycle manufacturing and injection molding industries in Wangqingtuo Town) and dense residential areas, while Funnel I-2 is primarily composed of residential areas. The high subsidence rates in these regions are attributed to long-term over-exploitation of groundwater for industrial production and domestic water use, which is consistent with previous research findings on groundwater-driven subsidence in the contiguous Tianjin-Wuqing-Bazhou area2,57. Funnel I-1 showed a tendency to connect into an elliptical feature oriented in the northeast-southwest direction, which reflected the lateral expansion of groundwater depression cones and the contiguous development of subsidence. Furthermore, the spatial overlap between Funnel IV-1 and the groundwater depression cone (Yangjiapo Town, Binhai New Area) further confirms that groundwater over-exploitation remains the primary driver of land subsidence in Tianjin. Subsidence Zone II and Subsidence Zone III exhibit moderate subsidence intensity but differ in their driving factors. For Subsidence Zone II, the maximum subsidence rates of Funnel II-1 (Wangwenzhuang Town in Xiqing District) and Funnel II-2 (Balitai Town in Jinnan District) are 63.26 mm/yr and 60.05 mm/yr, respectively. Funnel II-1 is located in a mixed industrial-residential area, where groundwater extraction for industrial cooling and domestic water use is the primary driver of subsidence; Funnel II-2 is surrounded by farmland and factories, with agricultural irrigation (especially during dry seasons) and industrial water use jointly contributing to subsidence58. In contrast, the maximum subsidence rates of Funnel III-1 (northern Junliangcheng in Dongli District) and Funnel III-2 (Airport Economic Zone) in Subsidence Zone III are relatively lower, at 45.15 mm/yr and 35.96 mm/yr respectively. This may be attributed to the construction of large-scale residential and industrial areas, as well as the influx of new residents, which has increased reliance on groundwater. Notably, with the development of the regional economy, the subsidence center in Tianjin has historically experienced shifts3. Different from previous studies18,50,57,58, Subsidence Zone III is a newly identified subsidence area in this research. We argue that this phenomenon is closely related to the construction of numerous new projects in the Airport Economic Zone in recent years. Large-scale construction activities, foundation pit dewatering, and the construction of dense high-rise buildings have led to increased groundwater extraction volume and greater extraction depth in local areas. This aligns with the broader consensus that the interaction of multiple factors (groundwater extraction, engineering activities, and geological conditions) exerts an increasingly significant impact on subsidence patterns in rapidly urbanizing regions. Spatial contradiction between GWS recovery and land subsidence The spatial inconsistency between positive GWS changes in Jinghai, Xiqing, and Jinnan districts and severe land subsidence in Zones I and II may be closely linked to the lagged compaction of clayey aquitards induced by historical deep groundwater over-exploitation, with three intertwined mechanisms potentially contributing to this phenomenon: First, Tianjin’s aquifer system exhibits distinct vertical stratification, separated by a 30–80 m thick clayey aquitard into a shallow unconfined aquifer (burial depth < 50 m) and deep confined aquifers (100–430 m)13. The positive GWS detected by GRACE is likely attributed to shallow recharge from the SNWDP-MRP, which passes through Xiqing and adjacent areas, while no large-scale artificial replenishment has been observed to reach deep aquifers55, indicating that deep groundwater depletion has not been fundamentally alleviated. Second, long-term deep groundwater over-exploitation since the 1980s may have triggered irreversible secondary consolidation (creep) of the clayey aquitard, which account for 28.3% of total deformation in Tianjin’s Late Pleistocene strata with a lag time possibly exceeding 25 years13. Such compaction might persist due to slow pore water pressure dissipation and soil creep even with reduced deep extraction2,48, meaning the historical compaction of aquitards cannot be reversed by short-term shallow GWS recovery. Third, Tianjin’s subsidence appears to be dominated by strata below 300 m (contribution rate 65%, over 70% in Xiqing and adjacent areas), which correspond to the historically over-exploited deep aquifer groups13. The spatial overlap of these deep subsidence-prone strata with shallow GWS recovery areas, combined with the long-term lagged compaction of aquitards, ultimately leads to the coexistence of positive shallow GWS changes and severe deep land subsidence. This finding implies that the correlation identified in this study may largely reflect the delayed response of land subsidence to historical groundwater drawdowns, rather than a simple direct relationship with concurrent GWS changes. The lagged compaction of aquitards driven by past over-exploitation remains a key controlling factor for current subsidence in Tianjin, even amid regional shallow GWS recovery. Lag effect between land subsidence and GWS changes The lag effect between land subsidence and GWS changes in Tianjin is a comprehensive manifestation of the combined effects of the hydro-mechanical properties of aquitards, initial hydrogeological conditions, and human activities. Essentially, this lag effect is closely associated with the hydro-mechanical characteristics of aquifer systems and the dynamic response process of soil deformation to groundwater level fluctuations. From the perspective of time-series characteristics, the cumulative subsidence curves of all seven subsidence funnels exhibit a consistent pattern of “fluctuating with GWS changes but with a time delay”. Taking Subsidence Zone I as an example, the cumulative subsidence of Funnel I-1 continued to increase from January 2023 to June 2024, while the growth rate slowed significantly after July 2024. This deceleration process lagged behind the start of GWS recovery in the region by approximately 6 months. In contrast, Funnel IV-1 in Subsidence Zone IV showed no obvious tendency of subsidence deceleration during the monitoring period, which was synchronized with the continuous decline of regional GWS. Quantitative verification via time-lagged cross-correlation further confirmed that the lag time of land subsidence relative to GWS changes across the four subsidence zones ranged from 41 to 263 days, with an average of 112 days (Table 2). Among them, Subsidence Zone II had the shortest lag time (41–112 days), and Subsidence Zone III had the longest (81–263 days), reflecting differences in the sensitivity of various subsidence zones to GWS changes. Another finding is that there is a positive correlation between the lag time and the correlation coefficient between cumulative subsidence and GWS changes (Table 2): Subsidence Zone I (with a lag time of 97–186 days) had a relatively high correlation coefficient (0.542–0.853), indicating a strong correlation between GWS changes and subsidence; in contrast, Subsidence Zone III (with a lag time of 81–263 days) had a lower correlation coefficient (0.257–0.882). This reflected that non-hydrological factors such as foundation pit dewatering in the Airport Economic Zone have weakened the direct correlation between GWS and subsidence, and the interaction of multiple factors in rapidly urbanizing areas has complicated the lag relationship between groundwater and subsidence. In summary, this finding provides certain reference significance for subsidence prediction and prevention: Regions with short lag times and high cross-correlation should prioritize the development of real-time groundwater monitoring, actively seek alternative water sources, and strictly restrict groundwater extraction; regions with long lag times and low cross-correlation need to strengthen research on the driving mechanisms of land subsidence to avoid the risk of cumulative deformation. In addition, it is recommended to incorporate the lag time parameter into land subsidence prediction models to improve the accuracy of early warnings. Limitations and future perspectives This study integrates PS-InSAR, GRACE, and GLDAS data to explore the relationship between land subsidence and GWS changes in Tianjin, focusing on their spatiotemporal characteristics and lag effects. Key limitations include: PS-InSAR measurements inherently suffer from various uncertainties, including atmospheric noise, phase unwrapping errors, and spatial resolution constraints, all of which may degrade the reliability of the retrieved deformation time series. In addition, the unavailability of field leveling observations for validation introduces further quantitative uncertainties into the absolute magnitudes of InSAR-derived land subsidence. PS-InSAR measurements inherently suffer from various uncertainties, including atmospheric noise, phase unwrapping errors, and spatial resolution constraints, all of which may degrade the reliability of the retrieved deformation time series. In addition, the unavailability of field leveling observations for validation introduces further quantitative uncertainties into the absolute magnitudes of InSAR-derived land subsidence. Significant spatial resolution mismatch: GRACE mascons data only captures large-scale regional GWS, and is physically incapable of resolving small-scale groundwater cones of depression that drive localized subsidence funnels; Significant spatial resolution mismatch: GRACE mascons data only captures large-scale regional GWS, and is physically incapable of resolving small-scale groundwater cones of depression that drive localized subsidence funnels; GRACE-derived GWS reflects the average variation of regional terrestrial water storage, which may fail to distinguish vertical groundwater stratification (e.g., shallow vs. deep aquifers) or the spatial heterogeneity of local groundwater extraction intensity. This may largely account for the phenomenon of overall regional GWS increase alongside severe subsidence in local hotspots (e.g., Subsidence Zones I and II in Jinghai, Xiqing, and Jinnan districts), and potentially weakens the accuracy of lag time calculations in small-scale areas. GRACE-derived GWS reflects the average variation of regional terrestrial water storage, which may fail to distinguish vertical groundwater stratification (e.g., shallow vs. deep aquifers) or the spatial heterogeneity of local groundwater extraction intensity. This may largely account for the phenomenon of overall regional GWS increase alongside severe subsidence in local hotspots (e.g., Subsidence Zones I and II in Jinghai, Xiqing, and Jinnan districts), and potentially weakens the accuracy of lag time calculations in small-scale areas. Land subsidence is a complex process governed by multiple natural and anthropogenic factors. In recent years, several studies have incorporated meteorological variables into relevant analyses; for instance, Ghaderpour et al.21 observed in the Catania Plain that, in addition to groundwater effects, temperature and precipitation also show significant annual coherency with surface deformation, with a 5–6 month phase lag detected in some areas. Together, these findings indicate that land subsidence generally exhibits delayed responses to external hydrological and climatic drivers, and the absence of meteorological factors in the present framework may introduce additional uncertainty in the comprehensive interpretation of lag mechanisms. Land subsidence is a complex process governed by multiple natural and anthropogenic factors. In recent years, several studies have incorporated meteorological variables into relevant analyses; for instance, Ghaderpour et al.21 observed in the Catania Plain that, in addition to groundwater effects, temperature and precipitation also show significant annual coherency with surface deformation, with a 5–6 month phase lag detected in some areas. Together, these findings indicate that land subsidence generally exhibits delayed responses to external hydrological and climatic drivers, and the absence of meteorological factors in the present framework may introduce additional uncertainty in the comprehensive interpretation of lag mechanisms. Future research will focus on three key aspects to address the above limitations: Collect and collate long-term field leveling and high-precision GNSS data to verify and calibrate the absolute accuracy of InSAR-derived subsidence results, further reducing quantitative uncertainties; Collect and collate long-term field leveling and high-precision GNSS data to verify and calibrate the absolute accuracy of InSAR-derived subsidence results, further reducing quantitative uncertainties; Adopt downscaling techniques such as geographically weighted regression (GWR) and random forest to downscale GRACE data59,60, realizing precise spatial matching between GWS changes and localized subsidence funnels. Adopt downscaling techniques such as geographically weighted regression (GWR) and random forest to downscale GRACE data59,60, realizing precise spatial matching between GWS changes and localized subsidence funnels. Explore additional methods to verify and supplement the lagged response results obtained in this study, such as the Least-Squares Cross-Wavelet Analysis (LSCWA) and cross-spectral analysis20,21, which can identify coherency, phase delay, and multi-timescale correlations between the relevant time series. Further explore the intrinsic response mechanisms between GWS changes (especially deep aquifer dynamics) and land subsidence, incorporate the identified groundwater-subsidence lag time as a key parameter into prediction models, and optimize model structures by introducing deep learning algorithms to improve the accuracy and reliability of long-term subsidence predictions. Explore additional methods to verify and supplement the lagged response results obtained in this study, such as the Least-Squares Cross-Wavelet Analysis (LSCWA) and cross-spectral analysis20,21, which can identify coherency, phase delay, and multi-timescale correlations between the relevant time series. Further explore the intrinsic response mechanisms between GWS changes (especially deep aquifer dynamics) and land subsidence, incorporate the identified groundwater-subsidence lag time as a key parameter into prediction models, and optimize model structures by introducing deep learning algorithms to improve the accuracy and reliability of long-term subsidence predictions. Incorporate key meteorological variables (e.g., precipitation, temperature) into the analytical framework to further examine the combined climate–groundwater controls on subsidence lag behavior, explore potential phase delays between meteorological factors and subsidence, and allow a more integrated understanding of deformation dynamics. Incorporate key meteorological variables (e.g., precipitation, temperature) into the analytical framework to further examine the combined climate–groundwater controls on subsidence lag behavior, explore potential phase delays between meteorological factors and subsidence, and allow a more integrated understanding of deformation dynamics. Conclusions Based on the latest 2023–2025 monitoring data, this study obtains several novel findings for Tianjin’s land subsidence: the identification of a previously overlooked subsidence zone (Zone III in Dongli District), the quantification of differentiated lag time characteristics of land subsidence to groundwater changes across four typical subsidence zones, and the discussion of the spatial contradiction between shallow GWS recovery and deep land subsidence driven by aquitard lagged compaction. These findings supplement the existing macro-scale research and provide updated fine-scale scientific support for Tianjin’s current subsidence prevention and control. This study obtained data on land subsidence and groundwater changes in the study area using PS-InSAR technology and GRACE time-variable gravity field inversion data. The reliability of the experimental data was verified using SBAS-InSAR technology and in-situ groundwater data. Finally, time-lagged cross-correlation analysis was applied to calculate the lag time of land subsidence relative to groundwater changes and the corresponding cross-correlation coefficients. The main conclusions are as follows: Reliability Verification of Experimental Data Reliability Verification of Experimental Data Cross-validation of results from PS-InSAR and SBAS-InSAR showed a linear regression R² of 0.86, with 95% of data point differences concentrated within the range of -1.19 to 9.78 mm/yr, confirming the accuracy of the land subsidence monitoring data. Additionally, direct quantitative validation using in-situ groundwater level data revealed significant consistency with GRACE-inverted GWS, with Pearson correlation coefficients ranging from 0.638 to 0.861 across five monitoring wells. 2.Spatial Heterogeneity of Land Subsidence and GWS Changes Spatial Heterogeneity of Land Subsidence and GWS Changes Severe land subsidence in Tianjin exhibits a distinct “peripheral aggregation” pattern. A total of 4 contiguous subsidence zones and 7 subsidence funnels were identified, primarily distributed in the periphery of the main urban area rather than the core region. Significant differences in subsidence intensity were observed: the most severe subsidence occurred in Zone I (centered on Wangqingtuo Town and Shengfang Town, with a maximum rate of 60.05 mm/yr) and Zone IV (near Yangjiapo Town, with a maximum rate of 67.62 mm/yr), while Zone III experienced relatively mild subsidence (maximum rate of 45.15 mm/yr). In contrast, GWS showed positive changes in most parts of the study area (with a maximum annual change rate of 59.13 mm/yr, mainly concentrated in the east-central part of Jinghai District, south-central part of Xiqing District, Jinnan District, and central part of Binhai New Area), decreasing gradually from the central region to the north and south. This spatial pattern is consistent with the peripheral aggregation in the distribution of severe land subsidence. Only the northern part of Binhai New Area showed negative GWS changes (a significant decline area, with a minimum change rate of -9.22 mm/yr), and this area overlaps with Subsidence Zone IV, indirectly confirming that over-exploitation of groundwater remains the primary driver of land subsidence in this region. 3.Lagged Relationship Between Land Subsidence and GWS Changes Lagged Relationship Between Land Subsidence and GWS Changes Time-lagged cross-correlation analysis indicated that land subsidence in Tianjin lags behind GWS changes by 41 to 263 days (with an average of 112 days), showing significant spatial variations. The lag time was positively correlated with the cross-correlation coefficient between cumulative subsidence and GWS changes: Subsidence Zones I (lag time: 97–186 days, correlation coefficient: 0.542–0.853), II (lag time: 41–112 days, correlation coefficient: 0.431–0.826), and IV (lag time: 69–159 days, correlation coefficient: 0.538–0.846), which had shorter lag times, exhibited stronger correlations. This reflected a more direct response of subsidence to GWS fluctuations. In contrast, Subsidence Zone III had a lower correlation coefficient (lag time: 81–263 days, correlation coefficient:0.257–0.882), possibly due to non-hydrological factors such as rapid urbanization in the Airport Economic Zone (e.g., foundation pit dewatering and large-scale construction activities), which weakened the groundwater-subsidence linkage. The lagged relationship identified in this study provides targeted guidance for land subsidence prevention and control. Regions with short lag times and high correlation coefficients (e.g., Zones I, II, and IV) should prioritize the establishment of real-time groundwater monitoring systems, expand alternative water sources, and strictly restrict groundwater extraction. For regions with long lag times and low correlation coefficients (e.g., Zone III), in-depth research on the multi-factor driving mechanisms (including engineering activities and geological conditions) is required to avoid the risk of cumulative deformation. It is recommended to incorporate lag time as a key parameter into land subsidence prediction models to improve early warning accuracy, thereby providing scientific support for groundwater resource management and geological disaster prevention in Tianjin. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements We would like to thank the anonymous reviewers for their constructive suggestions and comments. Author contributions Conceptualization, Y.L.; methodology, Y.L., W.Z., K.M.; writing, Y.L., W.Z.; visualization, Y.L., W.J., H.W.; funding acquisition, D.Z. Funding This research was funded by Innovation Base for Natural Resources Monitoring Technology in the Lower Reaches of Yongding River, Geological Society of China (grant number: KY02202504) and the geological survey projects of the China Geological Survey Bureau (grant number: DD202607102703). Data availability All data generated or analysed during this study are included in this published article. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement All data generated or analysed during this study are included in this published article.