Explore the vast range of titles available.

As Deep Neural Networks (DNNs) are being increasingly employed to make important simulations in rainfall‐runoff contexts, the demand for interpretability is increasing in the hydrology community. Interpretability is not just a scientific question, but rather knowing where the models fall flat, how to fix them, and how to explain their outcomes to scientific communities so that everyone understands how the model arrives at specific simulations This paper addresses these challenges by deciphering interpretable probabilistic DNNs utilizing the Deep Autoregressive Recurrent (DeepAR) and Temporal Fusion Transformer (TFT) for daily streamflow simulation across the continental United States (CONUS). We benchmarked TFT and DeepAR against conceptual to physics‐based hydrologic models. In this setting, catchment physical attributes were incorporated into the training process to create physics‐guided TFT and DeepAR configurations. Our proposed physics‐guided configurations are also designed to aggregate the patterns across the entire data set, analyze the sensitivity of key catchment physical attributes and facilitate the interpretability of temporal dynamics in rainfall‐runoff generation mechanisms. To assess the uncertainty, the modeling configurations were coupled with a quantile regression by adding Gaussian noise N(0,σ)$N\\,(0,\\sigma )$with increasing standard deviation to the individual catchment attributes. Analysis suggested that the physics‐guided TFT was superior in predicting daily streamflow compared to the original TFT and DeepAR as well as benchmark hydrologic models. Predictive uncertainty intervals effectively bracketed most of the observational data by simultaneous simulation of various percentiles (e.g., 10th, 50th, and 90th). Interpretable physics‐guided TFT proved to be a strong candidate for CONUS daily streamflow simulations.

Blue water (BW) and green water (GW) resources are fundamental to the hydrological cycle, yet their effective future projection and mechanistic understanding remain challenging. Both physics‐based and data‐driven models have limitations in complex hydrological long‐term projection and nonlinear interpretation. Therefore, an Interpretable Physics‐Guided Deep Learning (IPGDL) framework was proposed, which not only preserves the hydrological physical mechanisms of the Soil and Water Assessment Tool (SWAT) but also leverages the efficient nonlinear learning capabilities of CVOA‐CNN‐BiLSTM (CCB), while incorporating the Shapley Additive Explanations (SHAP) method to enhance result interpretability. Quality‐assured CMIP6 climate and PLUS land use scenarios drive the physics‐guided CCB model to project blue‐green water (2022–2100), while SHAP interprets nonlinear mechanisms through global, interaction, time step, and spatial effects from historical and future scenario perspectives. The IPGDL framework was applied to the Xiangjiang River Basin (XRB), and key findings are as follows: (a) The framework achieves robust projections by coupling physical mechanisms with nonlinear processes, reducing potential biases in traditional approaches. (b) Nonlinear interpretation results reveal meteorological features dominate blue‐green water (46% and 53% importance), with critical threshold effects (e.g., precipitation thresholds of 140.5 and 109.4 mm for BW and GW), significant temporal lags and spatial effects governing nonlinear system behavior in the historical period (1991–2020). Future scenario interpretation results reveal that low‐forcing scenarios due to precipitation‐dominated mechanisms trigger more extreme responses (−43.6% BW, +34.8% GW) with smaller uncertainty compared to medium and high‐forcing scenarios. This IPGDL framework demonstrates potential for hydrological modeling and offers insights into watershed management.

Accurate aerosol optical depth (AOD) prediction remains challenging due to complex aerosol‐radiation interactions and highly variable spatio‐temporal patterns. Three critical scientific issues motivate this work: understanding whether and how physical principles can enhance deep learning predictions, identifying which aerosol properties most strongly govern AOD variations, and improving the prediction of extreme AOD events critical for air quality management. Herein, utilizing MERRA‐2 reanalysis data (1980–2024) over the Huaihe River Basin in eastern China, a Physics‐Guided deep learning framework is presented for Aerosol Optical Depth (AOD) prediction. The model proposed integrates Convolutional Neural Networks (CNN), Long Short‐TermMemory (LSTM) networks, and multi‐head attention mechanisms to capture both spatio‐temporal features and physical relationships of aerosol properties. Three key aspects are involved: First, a hybrid deep learning model is developed and evaluated, which combines CNNs for spatial correlation extraction, bidirectional LSTM for temporal dependency modeling, and multi‐head attention for feature interaction learning. Second, a comprehensive feature importance analysis is conducted by examining the relationships between different aerosol properties (mass concentration, scattering coefficient, and Ångström exponent) and AOD prediction, offering physical insights into the model's decision‐making process. Third, a specialized approach is proposed for extreme AOD event prediction, focusing on early detection and accurate forecasting of high‐AOD episodes. Overall, the results demonstrate the model's efficacy in capturing both regular AOD variations and extreme events, with the Physics‐Guided architecture showing superior performance compared to traditional methods. This integrated approach enhances AOD prediction accuracy and deepens insights into aerosol‐radiation interactions, thereby improving atmospheric monitoring and air quality forecasting. While MERRA‐2 has inherent temporal delays, this framework provides valuable capabilities for historical trend analysis, numerical model validation, and can be readily adapted for real‐time applications through transfer learning with satellite observations. Plain Language Summary Aerosol Optical Depth (AOD) is a crucial measure of how much sunlight is blocked by particles in the air, affecting both climate and air quality. Traditional methods for predicting AOD often struggle with accuracy and efficiency. This study develops a new artificial intelligence model that combines physical principles with deep learning techniques to predict AOD over the Huaihe River Basin in eastern China. Our model shows significant improvements in prediction accuracy, particularly in identifying extreme pollution events. By analyzing different types of aerosol properties, we found that the way particles scatter light is more important for predictions than their mass. The model performs better in summer than in winter, likely due to winter's more complex weather conditions. This improved prediction system could help better forecast air quality and provide early warnings for severe pollution events, benefiting public health and environmental management. Key Points A hybrid deep learning model is developed and evaluated, for feature interaction learning A comprehensive feature importance analysis is conducted by examining the relationships between different aerosol properties A specialized approach is proposed for extreme AOD event prediction

Multistage centrifugal pumps (MCPs) are critical components in industrial systems, where early and reliable fault diagnosis remains challenging due to nonstationary operating conditions, noise contamination, and limited fault sensitive information in single domain representations. To address these issues, this paper proposes a physics guided fused (PGF) image learning framework with enhanced squeeze excitation (ESE) attention, for intelligent MCP fault diagnosis. First, a physics guided window selection strategy identifies the most informative signal segments by jointly considering energy concentration, impulsiveness, and fault related frequency band characteristics. From each selected segment, a PGF image is constructed by integrating a physics guided Mel spectrogram, a Gramian Angular Difference Field (GADF), and a Cross Interaction Map (CIM) that explicitly models their mutual dependency. This fused image captures complementary time frequency, nonlinear temporal, and interaction level fault characteristics in a unified representation. In addition, a low dimensional physics feature vector is extracted from each signal segment and injected into an ESE attention mechanism to adaptively recalibrate convolutional feature responses based on physical signal behavior. The proposed framework is validated on a real industrial MCP dataset under three operating pressures of 3 bar, 3.5 bar, and 4 bar, covering multiple fault conditions. Experimental results demonstrate consistently high diagnostic performance across all pressure levels, achieving accuracy of greater than 99% across all pressure bars with macro average F1 scores exceeding 0.99. These results confirm the robustness and generalization capability of the proposed physics guided fused image and attention learning framework for real world MCP fault diagnosis.

Response latency and sensor noise are universal challenges in closed-loop control systems. In the context of semi-active suspensions, these issues also exist and manifest as critical bottlenecks. Due to the highly transient nature of road shocks, the inherent physical actuation delays of the hardware, combined with the phase lag introduced by traditional signal filtering, often cause the control response to significantly lag behind the physical excitation. To address this issue from a predictive perspective, this study proposes a Physics-Informed Gated Convolutional Neural Network (PI-GCNN) designed to predict future multi-modal energy evolution, thereby enabling feedforward control. Unlike traditional feedback mechanisms, the proposed framework employs the Continuous Wavelet Transform (CWT) to convert short-horizon inertial data into time–frequency scalograms, effectively isolating transient shock features from background vibrations. A novel physics-guided gating mechanism is embedded within the network architecture to regulate feature activation. This mechanism is trained using an asymmetric sparse physics loss, which combines L1 regularization with adaptive spectral consistency constraints to enforce noise suppression on flat roads while ensuring sensitivity to impacts. Extensive validation was conducted using high-fidelity heavy truck simulations and the public PVS 9 real-world dataset. The results confirm that the PI-GCNN achieves a predictive phase lead of approximately 100–200 ms over real-time baselines, creating a valuable actuation window for suspension dampers. Furthermore, the model demonstrates exceptional computational efficiency, with a parameter count of 0.10 M and a single-frame inference latency of 0.25 ms, making it highly suitable for deployment on resource-constrained automotive edge computing platforms.

Urban flooding is intensifying under climate change and urbanization, demanding efficient deep learning-based prediction. However, such models are commonly trained to minimize data-fitting loss alone, with limited incorporation of physical constraints on surface water flow. As a result, they may learn spurious statistical relationships and provide little insight into the factors governing predicted inundation, limiting their practical value for flood risk management. This study develops a physics-guided and interpretable deep learning framework for compound rainfall-tidal flood prediction in a representative coastal island setting, where flood dynamics are strongly shaped by interactions between rainfall forcing, terrain controls, and tidal boundary conditions. The framework integrates a U-Net encoder-decoder for spatial feature extraction, along with Bidirectional LSTM branches and multi-head self-attention, to encode rainfall and tidal time series. Physics-guided loss functions that enforce gradient consistency and spatial smoothness are introduced via staged weight scheduling to improve physical plausibility while preserving predictive accuracy. Model interpretability is further achieved using Integrated Gradients to quantify feature contributions, with robustness confirmed by K-fold stability analysis and independent ablation experiments. Results show that drainage junctions emerge as the dominant predictor under the present architecture–task setting, while terrain-related variables jointly account for the majority of attribution, indicating that the model captures key hydraulic and topographic controls on inundation. The resulting importance hierarchy, in which drainage junctions and geomorphological features jointly exceed the contribution of elevation, is preserved in a paired no-physics baseline trained under an otherwise identical configuration, with a cross-model Spearman rank correlation of ρ = 1.00 under the mean-depth target. This robustness identifies the hierarchy as a property of the architecture-task pairing rather than of the physics-guided loss terms; the latter act instead as an output-level hydraulic regulariser, reducing gradient-consistency violations by 16.7% at a marginal 0.51% reduction in R². These findings demonstrate that integrating physical constraints with gradient-based attribution analysis can yield more credible and interpretable deep learning predictions for compound urban flooding.

We present a Physics-guided deep learning framework to address common limitations in Confocal Laser Scanning Microscopy (CLSM), including diffraction-limited resolution, noise, and under sampling due to low laser power conditions. The optical system’s point spread function and primary CLSM image degradation mechanisms, namely photon shot noise, dark current noise, motion blur, speckle noise, and under sampling are explicitly incorporated into the model as physics-based constraints. A convolutional autoencoder is trained with a custom loss function that integrates these optical degradation processes, ensuring that the reconstructed images adhere to physical image formation principles. The model is evaluated on simulated CLSM datasets generated based on experimentally observed CLSM noise characteristics. Statistical comparisons, including intensity histograms, spatial frequency distributions, and structural similarity metrics, confirm that the synthetic dataset closely matches accurate CLSM data. The proposed approach is compared with traditional image reconstruction methods, including Richardson-Lucy deconvolution, non-negative least squares, and total variation regularization. Results indicate that the physics-constrained autoencoder improves structural detail recovery while maintaining consistency with known CLSM imaging physics. This study demonstrates that Physics-guided deep learning can provide an alternative computational approach to CLSM enhancement, complementing existing optical correction methods. Future work will focus on further validation using experimental CLSM acquisitions.

Neural operators have been developed to learn a highly nonlinear mapping between input fields and solution fields of mechanics problems, achieving significant speedup compared to conventional solvers. However, the application of neural operators in solid mechanics problems is still limited. To solve both forward and inverse problems in structural engineering, this study develops a two‐dimensional vehicle–bridge Interaction Neural Operator (VINO2D) framework, achieving efficient bridge dynamic simulations and real‐time damage detection on numerical datasets. VINO2D learns the mapping between the damage distribution field and the two‐dimensional solution field of bridge response as a function of coordinates and time. A finite element (FE) simulation dataset of a bridge structure with an arbitrary damage distribution field is established to train the Fourier neural operator (FNO) model to learn the complicated nonlinear mappings between input damage distribution fields and structural response fields. The results indicate that forward VINO2D models achieve high‐fidelity simulation of bridge dynamics responses induced by a traveling vehicle with a speedup of more than 2000 times compared to conventional FE solvers. Furthermore, the inverse VINO2D model demonstrates high accuracy with 5%–7% errors in the real‐time inference of the damage distribution field along the bridge span from the structural response fields. The proposed method was validated through rigorous numerical experiments and may be combined with a state‐of‐the‐art computer vision algorithm to achieve real‐time model updating of bridge structures and digital twins.

Real-time processing capability is a key enabler for the next generation of intelligent satellites. Full-waveform spaceborne laser altimetry has become an indispensable tool in a wide range of scientific and engineering applications, including terrain mapping, biomass estimation, and Earth system monitoring. Within this context, accurate and efficient laser footprint localization is essential for ensuring the geometric reliability of altimetric measurements. However, conventional waveform-matching methods suffer from severe computational burdens, rendering them unsuitable for high-frequency on-orbit calibration and difficult to deploy in real-time onboard scenarios. To address these challenges, this paper proposes SLA-FLNet – a physics-guided deep learning framework that integrates key physical mechanisms of laser pulse propagation, terrain modulation, and echo formation through a multi-branch spatiotemporal architecture. Each module of SLA-FLNet explicitly encodes a physically interpretable process, enabling accurate, interpretable, and scalable footprint localization. To support supervised training in the absence of ground-truth labels, pseudo-labels were generated using a classical waveform-matching algorithm. The model was evaluated on 2379 laser footprints from 18 beams of the GaoFen-7 (GF-7) satellite, spanning 12 U.S. states with diverse terrain. SLA-FLNet achieved high prediction accuracy and delivered footprint localization results consistent with waveform matching, even in unseen geographic regions. An ablation study further highlighted the critical role of terrain-encoding in enhancing structural fidelity and cross-regional generalization. Compared to traditional methods, SLA-FLNet achieved over 100,000 × inference speedup on modern GPUs, demonstrating strong potential for real-time onboard processing. In summary, SLA-FLNet provides a physically consistent, computationally efficient, and deployment-ready solution for full-waveform footprint localization and on-orbit calibration, supporting future autonomous Earth observation missions.

Noninvasive imaging through scattering media is crucial for diverse applications but remains constrained by a narrow field of view (FOV). Although recent learning‐based methods have a larger FOV, they often require large‐scale real experimental datasets and struggle when the FOV is far beyond the optical memory effect (OME). Here, we propose a physics‐guided adaptive dual‐domain diffusion model for ultra‐wide‐field noninvasive imaging through scattering media, namely UNI‐Net. Specifically, we first develop a physical scattering imaging model to synthesize large‐scale pre‐training data, thereby reducing dependence on real experimental datasets. Second, to maximize the utilization of speckle information, we partition each speckle pattern into multi‐channel patches to guide the diffusion process. Third, we propose a spatial‐channel parallel attention block to model the spatial sparsity and inter‐channel similarity of speckle patches with linear complexity. Extensive experiments show that our method cuts reliance on real experimental data by an order of magnitude and achieves a PSNR of 31.23 dB at a 41 OME range in complex scenes, which is 49.5% higher than existing approaches while requiring significantly lower computational and memory costs. Even at an extreme 164 OME range where other methods fail, it still reliably reconstructs complex scenes with a PSNR of 27.21 dB.