Peri-urban floodplains include a diverse range of potential contaminant sources and complex surface water-groundwater interactions which control the transport of emerging contaminants in underlying groundwater systems. The impact of an unlined legacy landfill located within the floodplain of the River Thames in southern United Kingdom is assessed with two downgradient multi-depth piezometer transects, and results compared with a control piezometer site previously shown to be unaffected by the landfill leachate plume. Groundwater samples collected from two monitoring rounds in 2017 and 2018 from a network of sites showed the landfill to be a heterogeneous source of contaminants. Over 580 micro-organics (MO) including pharmaceuticals, pesticides, lifestyle products, surfactants and per- and polyfluoroalkyl substances (PFAS) were screened in groundwater and surface water using an semi-quantitative LCMS screen. Use of sucralose (an artificial sweetener) post-dates the closure of the landfill. Sucralose found within the landfill plume, together with contaminants found within the control site, suggests a more recent source of contamination also contributes to groundwater contamination within the floodplain. As these samples do not match the fingerprint of the River Thames, it is proposed that the additional contaminant source could be infiltration from either a canal or train wastewater discharge upgradient of the landfill. Spatial variation shows contaminant attenuation down gradient of the landfill with higher concentrations found in the deeper part of the alluvial gravel aquifer system possibly due to local recharge diluting shallow groundwater. Although closed 40 years prior to this study, groundwater sampling has demonstrated that the landfill continues to be a source of persistent, mobile contaminants.
Monitoring excess pore water pressure (EPWP) during and after soil improvement is essential under certain ground conditions. In some cases, ground monitoring may be required throughout the service life of a structure, particularly in areas where the subsoil is susceptible to environmental changes. However, installing sensors at every location across a construction site and its surrounding area is often impractical due to various site-specific constraints. This study proposes a spatial prediction framework that integrates B-spline interpolation and artificial neural networks (ANNs) to estimate EPWP at uninstrumented locations using simultaneous measurements from existing sensors. Experimental data reported in the published literature are extracted and processed in MATLAB to develop the ANN model and integrates results by B-spline interpolation. The network architecture is appropriately configured, and the available data are divided into training, validation, and testing subsets to evaluate predictive performance and reduce the risk of overfitting. The results indicate good agreement between the ANN predicted EPWP (kPa) at uninstrumented location and the experimental results reported in the literature. The proposed framework aims to apply these findings to EPWP monitoring and liquefaction assessment during and after soil improvement.
BackgroundEarthquake-induced cascading damage and functional failure of dense building clusters are the primary source of systemic disaster losses in megacities, yet conventional regional seismic assessments generally neglect spatial variations in local site conditions, which misrepresents ground-motion input and the resulting resilience pattern.MethodsUsing the time-averaged shear-wave velocity of the uppermost 30 m (Vs30) as the core site metric, spatially variable ground motions were refined through site response analysis, explicitly accounting for basin amplification, while soil–structure interaction was addressed in the refined single-building analyses. National seismic risk survey data were integrated with screened specialized datasets to construct an optimized GIS database and a three-tier “region–system–individual building” evaluation architecture; a three-dimensional indicator system covering site geology, structural safety and post-earthquake functional recovery was then applied to a core district of Beijing, China.ResultsVs30-based site correction captured the spatial heterogeneity of ground motions and substantially altered the ground-motion intensity fields, raising peak ground acceleration from a baseline of 3.80 m/s2 to 5.80 m/s2 in the southeastern soft-soil subdistricts. Severely damaged or fully collapsed buildings accounted for 26.46% of the inventory, whereas 95.09% of buildings complied with the predefined resilience targets, the 4.91% that failed being concentrated in historic aged residential subdistricts overlapping thick soft soils. Refined single-building analyses ranked resilience as base-isolated RC shear-wall residential > RC frame–shear wall office > non-isolated RC-frame hospital.ConclusionSite-corrected, building-level resilience mapping provides quantitative evidence for megacity seismic planning, old-community renewal and emergency resource deployment, and shows that addressing primary structural safety alone systematically underestimates the vulnerability of critical lifeline facilities.
Artificial intelligence (AI) is increasingly applied to infrastructure systems to improve efficiency, resilience, sustainability, and decision-making in rapidly changing, uncertain environments. This study presents a systematic literature review and analytical synthesis of AI applications in sustainable and resilient infrastructure systems, focusing on energy, water, transportation, and urban development. The review examined peer-reviewed studies and selected institutional reports identified primarily through Google Scholar, SpringerLink, ScienceDirect, and relevant international organizations, with emphasis on recent literature published between 2020 and 2026. The study screened retrieved studies and thematically categorized them by infrastructure sector, data sources and types, data-collection technologies, data infrastructure and platforms, AI methods, decision-support functions, sustainability and resilience outcomes, and implementation challenges. The synthesis shows that AI supports predictive monitoring, demand forecasting, anomaly detection, resource optimization, infrastructure management, and decision support across the four sectors. However, the evidence remains uneven across sectors and geographic contexts, with substantial gaps in long-term empirical validation, cross-sector integration, interoperable data ecosystems, lifecycle assessment, and applications in developing regions. The review also identifies challenges related to data quality and availability, digital infrastructure, interoperability, cybersecurity, institutional capacity, digital inequality, and the resource requirements associated with maintaining AI-enabled infrastructure systems. This review’s novel contribution is an integrated data–AI infrastructure–sustainability perspective that connects data sources and collection technologies with data infrastructure, AI methods, decision-support functions, and infrastructure outcomes. This perspective provides a basis for understanding how AI can move beyond isolated, sector-specific applications toward more integrated, adaptive, resilient, and innovative AI-driven solutions for sustainable infrastructure systems.
Unsupervised detection of bridge damages from vehicle-induced vibration is complicated by passage-to-passage variability and scarce damage labels. This study proposes a distribution-based framework using acceleration from one mid-span accelerometer, with vehicle entry/exit times and speed additionally required for event extraction and spatial alignment. For each passage, a dual-branch Conv–LSTM–attention autoencoder trained only on healthy data jointly encodes the spatially aligned time-domain response and the passage-level FFT magnitude spectrum. This fusion combines transient/local response information with global spectral characteristics. Frequency-domain reconstruction errors from multiple passages are treated as an empirical distribution, whose deviation from a healthy reference is quantified by the second-order Wasserstein distance. The framework is evaluated on a full-scale laboratory bridge under controlled but variable vehicle passages, with condition changes introduced by guardrail removal, and on a calibrated vehicle–bridge finite-element model with localized stiffness reductions. Compared with the frequency-only model examined herein, the fused model provides clearer separation between baseline and altered states. For the investigated cases, the index increases monotonically with imposed change severity, shows location-dependent sensitivity, and becomes less variable as passages accumulate, with approximate stabilization at 200–260 passages. The results support the feasibility of detecting global bridge condition changes using one bridge-mounted accelerometer plus auxiliary passage information, but do not establish damage localization or generalization to operational traffic.
