The increasing concentration of atmospheric CO2 since the Industrial Revolution has driven research into subsurface storage as a viable solution. This study focuses on developing machine learning models to estimate net present value and carbon footprint in a combined gas production and CO2 sequestration scenario in shales. The dataset comprised a large set of numerical simulation scenarios, which were run using PSU SHALECOMP, a 3-dimensional, compositional and multiphase simulator, which incorporates an equation of state to capture the effects of pressure and temperature variations. A horizontal production/injection well with multiple hydraulic fractures was modeled using the stimulated reservoir volume approach which represents the volume impacted by hydraulic fractures as well as the induced fractures through the natural fracture network in the reservoir. The results of these scenarios were used to calculate net present value and carbon footprint associated with each scenario. Exploratory data analysis and feature engineering revealed that the net present value is primarily governed by stimulated reservoir volume’s fracture permeability, original gas in place within the stimulated reservoir volume, and injection constraints, whereas the carbon footprint is predominantly controlled by total production duration and injected CO2 volume. Machine learning models were trained to build robust forecasting tools for net present value and carbon footprint. These models revealed that the selected neural network model outperformed multiple linear regression and random forests models in predicting both net present value and carbon footprint, with R2 values of 0.99 and 0.96, respectively, for the testing sets. To further refine these estimates and improve the robustness of predictions, future research should focus on improving the certainty in deterministic and probabilistic estimations of net present value and carbon footprint by gathering more comprehensive data, and conducting detailed analyses of carbon emissions and operational costs. This research represents a significant step toward understanding the economic and environmental implications of CO2 sequestration in shale reservoirs, contributing valuable insights for future developments in this field.
Living in a changing climate, human society now faces more frequent and severe natural disasters than before. As a consequence, rapid disaster response during the “Golden 72 Hours” of search and rescue becomes a vital humanitarian necessity and community concern. However, traditional disaster damage analysis routinely fails to generalize across distinct urban morphologies and new disaster events. Effective damage mapping typically requires exhaustive and time-consuming manual data annotation. To address this issue, we introduce Smart Transfer, a novel Geospatial Artificial Intelligence (GeoAI) framework, leveraging state-of-the-art vision Foundation Models (FMs) for rapid building damage mapping with post-earthquake Very High Resolution (VHR) imagery. Specifically, we design two novel model transfer strategies: first, Pixel-wise Clustering (PC), ensuring robust prototype-level global feature alignment; second, a Distance-Penalized Triplet (DPT), integrating patch-level spatial autocorrelation patterns by assigning stronger penalties to semantically inconsistent yet spatially adjacent patches. Extensive experiments and ablations on the recent 2023 Türkiye-Syria earthquake show promising performance in multiple cross-region transfer settings, namely Leave One Domain Out (LODO) and Specific Source Domain Combination (SSDC), achieving an average F1-score of 0.64 and mIoU of 0.69. Moreover, Smart Transfer provides a scalable, automated GeoAI solution to accelerate building damage mapping and support rapid disaster response, offering new opportunities to enhance disaster resilience in climate-vulnerable regions and communities. The data and code are publicly available at https://github.com/ai4city-hkust/SmartTransfer.
Abstract Substandard reinforced concrete (RC) building stocks continue to pose a major seismic risk in earthquake-prone regions. While fiber-reinforced polymer (FRP) jacketing has been widely demonstrated as an effective retrofitting technique at the member level, its rational application at a large scale requires rapid yet reliable decision-making tools. In this study, the accuracy and applicability of a rapid FRP retrofitting framework integrated into the PERA2019 performance-based seismic assessment methodology are evaluated through a detailed investigation of five actual representative existing RC buildings. For each building, seismic performance before and after FRP retrofitting was evaluated using both the PERA2019 procedure and detailed nonlinear pushover analyses, which served as benchmark solutions. In addition, the framework was validated against the results of a full-scale experimental study reported in the literature. The results indicate that PERA2019 provides seismic safety ratio estimates that are in satisfactory agreement with detailed analysis outcomes. Buildings classified as A class by PERA2019 were consistently shown to satisfy the corresponding controlled damage performance target in detailed analyses, whereas buildings failing to reach A class did not meet this target. A comparison with the experimental results further demonstrated that the framework accurately captured the relative performance improvement achieved through FRP retrofitting and remained on the safe side of the observed structural response. Overall, the findings confirm that the proposed methodology can reliably distinguish buildings that can be effectively upgraded using FRP jacketing from those requiring alternative interventions. The framework provides a computationally efficient and practically applicable tool for large-scale seismic risk mitigation and prioritization of FRP retrofitting strategies.
Smart-building load forecasters are often trained offline on dense, multivariate, high-frequency data, but deployment may provide only hourly, feature-limited inputs. Missing features must then be reconstructed, and their errors can propagate through the model. If this input uncertainty is not reflected, prediction intervals may become miscalibrated, affecting demand-response scheduling. Our work examines where uncertainty should be placed once inference inputs are reconstructed. We develop a unified one-day-ahead probabilistic forecasting framework that aligns temporal resolution, reconstructs the unavailable inputs, and derives causal features, and we compare a modular post-hoc residual-quantile scheme with an integrated in-model quantile-learning scheme. The comparison uses three mid-scale Deep Learning (DL) backbones: recurrent, hybrid recurrent, and attention-based Temporal Fusion Transformer (TFT) models, under identical inputs, forecasting horizon, preprocessing rules, and training budgets. Results show that uncertainty placement is backbone-dependent. Integrated quantile learning is most reliable with the TFT, yielding 2.2-3.6% MAPE and 28-83 W RMSE on the labeled test window, while producing intervals about 5 & times; narrower than the modular intervals at the closest-to-nominal coverage level. Diebold-Mariano tests support the TFT ranking and the mixed behavior of the recurrent backbones. A reconstruction-sensitivity test shows that reconstructed inputs increase the Quantile Score (QS) by 106% while interval width remains nearly unchanged, indicating that the model does not automatically absorb reconstruction-induced uncertainty. Robustness checks against non-DL baselines and seasonal hold-out weeks support this ranking. Our results expose the limits of post-hoc residual quantiles when inference depends on reconstructed inputs.
We report the first successful synthesis of a novel magnetically recoverable nanocatalyst, Fe3O4@Al2O3/Co-Cs, incorporating cesium (Cs) as a co-catalytic promoter for efficient and sustainable hydrogen generation via NaBH4 hydrolysis. In this system, magnetite (Fe3O4) serves as a magnetic core, enabling facile separation and recyclability, whereas the Al2O3 shell enhances thermal stability, structural integrity, and active-site dispersion. Cobalt (Co) acts as the primary active metal, whereas Cs is introduced here for the first time in NaBH4 hydrolysis as an electronic promoter, significantly enhancing electron transfer and accelerating the hydrogen generation rate (HGR). Comprehensive characterization using FTIR, XRD, FE-SEM/EDX, and BET analyses confirmed the successful formation of a hierarchical core-shell structure. FE-SEM images revealed a flower-like morphology composed of aggregated nanosheets and nanoparticles that promoted efficient mass transfer and gas diffusion. Catalytic performance tests conducted at 30 degrees C demonstrated a high hydrogen generation rate of 17.24 L g- 1 min- 1. The apparent activation energy was determined to be 27.18 kJ mol-1, indicating favorable reaction kinetics under mild conditions. Thermodynamic analysis revealed a low enthalpy of adsorption (Delta Hads = 35.66 +/- 0.01 kJ mol-1) and a small entropy change (Delta S degrees = 0.09022 +/- 0.01 kJ mol- 1 K-1), suggesting favorable interactions between the reactants and catalyst surface. Density functional theory (DFT) calculations further confirmed that the Fe3O4@Al2O3/Co-Cs system exhibited enhanced adsorption strength and reduced activation barriers, facilitating NaBH4 hydrolysis through efficient charge transfer, intermediate stabilization, and synergistic catalytic effects. Overall, this study highlights the novel role of cesium as an alkali metal promoter in NaBH4 hydrolysis, opening new opportunities for the design of advanced non-noble metal catalysts for hydrogen generation.
