Search

The existing machine learning (ML) models for vibration-based damage assessments often use highly compressed features or are restricted to preprocessed signals with fixed durations and sampling rates. Additionally, the learning capacities of ML models and computational resources are limited, which restricts using raw signals as direct input. This paper studied Mel filter banks (MFBs) for seismic signal processing, inspired by speech recognition technology. It is argued that the same filter designs in audio engineering may not be appropriate for seismic records, and therefore, a customized filter bank formulation was developed. Hybrid deep learning models for rapid assessments (HyDRA) were introduced as multibranch neural network architectures that enable end-to-end training for different types of processed vibration data structures. Moreover, the performance-based earthquake engineering (PBEE) equation was adjusted to integrate ML model uncertainties for probabilistic assessments. The proposed concepts were validated in a case study based on a data set of 32,400 nonlinear time-history analyses of a highway bridge in California. Insights and guidelines are provided for optimum filter design based on 5,184 experiments. Several HyDRA architectures were compared with benchmark models. The optimized MFB feature type outperformed features obtained from continuous wavelet transform and a stacked vector of conventional earthquake engineering indexes. A Bayesian variant of HyDRA was investigated to showcase its integration in the modified PBEE equation. Adopting custom filter banks with the HyDRA architecture enables effective feature extraction from raw vibration records by diversifying feature space and preserving information in the time and frequency domains. [ABSTRACT FROM AUTHOR]

Channel allocation has a direct and profound impact on the performance of vehicle-to-everything (V2X) networks. Considering the dynamic nature of vehicular environments, it is appealing to devise a blended strategy to perform effective resource sharing. In this paper, we exploit deep learning techniques predict vehicles’ mobility patterns. Then we propose an architecture consisting of centralized decision making and distributed channel allocation to maximize the spectrum efficiency of all vehicles involved. To achieve this, we leverage two deep reinforcement learning techniques, namely deep Q-network (DQN) and advantage actor-critic (A2C) techniques. In addition, given the time varying nature of the user mobility, we further incorporate the long short-term memory (LSTM) into DQN and A2C techniques. The combined system tracks user mobility, varying demands and channel conditions and adapt resource allocation dynamically. We verify the performance of the proposed methods through extensive simulations and prove the effectiveness of the proposed LSTM-DQN and LSTM-A2C algorithms using real data obtained from California state transportation department. [ABSTRACT FROM AUTHOR]

Traffic flow forecasting, which requires modelling involuted spatial and temporal dependence and uncertainty regarding road networks and traffic conditions, is a challenge for intelligent transportation systems (ITS). Recent studies have mainly focused on modelling spatial-temporal dependence through a fixed weighted graph based on prior knowledge. However, collecting up-to-date and accurate road information is costly. Moreover, is a single fixed graph enough to describe the correlation between sensors? The fixed weighted graph cannot directly relate to prediction tasks, which may result in considerable biases. To tackle this issue, in this paper, we propose a novel deep learning model framework: an adaptive spatial-temporal graph attention network (ASTGAT). Our ASTGAT simultaneously learns the dynamic graph structure and spatial-temporal dependency for traffic flow forecasting. Specifically, our framework consists of two joint training parts: a Network Generator model that generates a discrete graph with the Gumbel-Softmax technique and a Spatial-Temporal model that utilizes the generated network to predict traffic speed. Our Network Generator can adaptively infer the hidden correlations from data. Moreover, we propose a graph talking-heads attention layer (GTHA) for capturing spatial dependencies and design a gate temporal convolution (GTCN) layer for handling long temporal sequences. We evaluated our ASTGAT on two public datasets: METR-LA is collected in Los Angeles and PEMS-BAY is collected in California. Experimental results indicate that our ASTGAT outperforms the state-of-the-art (SOTA) baselines. Finally, to further describe our model, we visualize the forecasting results and the generated graph. [ABSTRACT FROM AUTHOR]

Wildfires have caused devastating consequences to property and human and animal lives, which has become a global problem. Consequently, advanced wildfire prediction models are required to treat complex features and climate conditions. As a result, Machine Learning and Deep Learning models are becoming popular. However, creating a balanced true and false labeled dataset in the wild-fire domain is often challenging. Hence, One-class classification models are a promising approach to overcome this concern. In this paper, several One-class classification models are investigated; linear models (Principal Component Analysis and One-Class Support Vector Machines), outlier ensemble models (Lightweight On-line Detector of Anomalies and Locally Selective Combination of Parallel Outlier Ensembles), proximity-based models (Histogram-based Outlier Score and Rotation-based Outlier Detection), probabilistic models (Unsupervised Outlier Detection Using Empirical Cumulative Distribution Functions and Copula-Based Outlier Detection), neural network-based models (Deep One-class Classification and Adversarially Learned Anomaly Detection) is used in two case studies for California and Western Australian states. In conclusion, it was found that Deep learning-based One-class classification models outperform other models in terms of performance and feature importance of showcasing the effectiveness of deep neural network models in the wildfire prediction domain. [ABSTRACT FROM AUTHOR]

The quality of traffic services depends on the accuracy and completeness of the collected traffic data. However,the existing traffic data imputation methods usually only rely on the predefined road network structure to capture the spatio-temporal features and only consider the imputation effect from a single perspective, which are very limited for imputation of different missing patterns of road traffic data. In this paper, we propose a novel deep learning framework called Hierarchical Spatio-temporal Graph Convolutional Neural Networks(HSTGCN) to impute traffic data,through the macro layer and the road layer. The model constructs macro graph of the road network based on the data temporal correlation clustering, which can mine the temporal dependencies of road traffic data from a hierarchical perspective. Besides, a temporal attention mechanism and adaptive adjacency matrix are introduced in the road layer to better extract the spatio-temporal information of the road traffic data. Finally, we use graph convolution neural networks to learn the spatio-temporal feature representations of the road layer and macro layer, which are then fused to achieve data imputation. To illustrate the efficient performance of the model, experiments are conducted on traffic data collected from California and Seattle. The proposed model performs better than the comparison model for traffic data imputation. • A full dynamic graph is built based on the urban road structure. • Extracting spatio-temporal features from multiple dimensions using hierarchical thinking. • Gate-GCN perform effective feature selection. [ABSTRACT FROM AUTHOR]

Distribution shifts remain a problem for the safe application of regulated medical AI systems, and may impact their real-world performance if undetected. Postmarket shifts can occur for example if algorithms developed on data from various acquisition settings and a heterogeneous population are predominantly applied in hospitals with lower quality data acquisition or other centre-specific acquisition factors, or where some ethnicities are over-represented. Therefore, distribution shift detection could be important for monitoring AI-based medical products during postmarket surveillance. We implemented and evaluated three deep-learning based shift detection techniques (classifier-based, deep kernel, and multiple univariate kolmogorov-smirnov tests) on simulated shifts in a dataset of 130'486 retinal images. We trained a deep learning classifier for diabetic retinopathy grading. We then simulated population shifts by changing the prevalence of patients' sex, ethnicity, and co-morbidities, and example acquisition shifts by changes in image quality. We observed classification subgroup performance disparities w.r.t. image quality, patient sex, ethnicity and co-morbidity presence. The sensitivity at detecting referable diabetic retinopathy ranged from 0.50 to 0.79 for different ethnicities. This motivates the need for detecting shifts after deployment. Classifier-based tests performed best overall, with perfect detection rates for quality and co-morbidity subgroup shifts at a sample size of 1000. It was the only method to detect shifts in patient sex, but required large sample sizes ( > 3 0 ′ 000 ). All methods identified easier-to-detect out-of-distribution shifts with small (≤300) sample sizes. We conclude that effective tools exist for detecting clinically relevant distribution shifts. In particular classifier-based tests can be easily implemented components in the post-market surveillance strategy of medical device manufacturers. [ABSTRACT FROM AUTHOR]

There are an estimated 800,000 small reservoirs globally with a range of uses. Given the collective importance of these reservoirs to water resource management and wider society, it is essential that we can monitor and understand the hydrological dynamics of ungauged reservoirs, particularly in a changing climate. However, unlike large reservoirs, continuous and systematic hydrological observations of small reservoirs are often unavailable. In response, this study has developed a retrieval framework for water levels of small reservoirs using a deep learning algorithm and remotely sensed satellite data. Demonstrated at four reservoirs in California, satellite imagery from both Sentinel-1 and Sentinel-2 along with corresponding water level field measurements was collected. Post-processed images were fed into a water level inversion convolutional neural network model for water level inversion, while different combinations of these satellite images, sampling approaches for training/testing data, and attention modules were used to train the model and evaluated for accuracy. The results show that random sampling of training data coupled with Sentinel-2 satellite imagery was generally the most accurate initially. Performance is improved by incorporating a channel attention mechanism, with the average R2 increasing by 8.6% and the average RMSE and MAE decreasing by 15.5% and 36.4%, respectively. The proposed framework was further validated on three additional reservoirs in different regions. In conclusion, the retrieval framework proposed in this study provides a stable and accurate methodology for water level estimation of small reservoirs and can be a powerful tool for small reservoir monitoring over large spatial scales. [ABSTRACT FROM AUTHOR]

The photovoltaic (PV) industry boom has accelerated the need for accurately understanding the spatial distribution of PV energy systems. The synergy of remote sensing and artificial intelligence presents significant prospects for PV energy monitoring. Currently, numerous studies have focused on extracting rooftop PV systems from airborne or satellite imagery, but their small-scale and size-varying characteristics make the segmentation results suffer from PV internal incompleteness and small PV omission. To address these issues, this study proposed a size-aware deep learning network called Rooftop PV Segmenter (RPS) for segmenting small-scale rooftop PV systems from high-resolution imagery. In detail, the RPS network introduced a Semantic Refinement Module (SRM) to sense size variations of PV panels and reconstruct high-resolution deep semantic features. Moreover, a Feature Aggregation Module (FAM) enhanced the representation of robust features by continuously aggregating deeper features into shallower ones. In the output stage, a Deep Supervised Fusion Module (DSFM) was employed to constrain and fuse the outputs at different scales to achieve more refined segmentation. The proposed RPS network was tested and shown to outperform other models in producing segmentation results closer to the ground truth, with the F1 score and IoU reaching 0.9186 and 0.8495 on the publicly available California Distributed Solar PV Array Dataset (C-DSPV Dataset), and 0.9608 and 0.9246 on the self-annotated Heilbronn Rooftop PV System Dataset (H-RPVS Dataset). This study has provided an effective solution for obtaining a refined small-scale energy distribution database. [ABSTRACT FROM AUTHOR]

Modeling the spread of wildland fires is essential for assessing and managing fire risks. However, this task remains challenging due to the partially stochastic nature of fire behavior and the limited availability of observational data with high spatial and temporal resolutions. Herein, we propose an attention-based deep learning modeling approach that can be used to learn the complex behaviors of wildfires across different fire-prone regions. We integrate optimized spatial and channel attention modules with a convolutional neural network (CNN) modeling architecture and train the attention-based fire spread models using a recently derived fire-tracking satellite observational dataset in conjunction with corresponding fuel, terrain, and weather conditions. The evaluation results and their comparison with benchmark models, such as a deeper and more complex autoencoder model and the semi-empirical FARSITE fire behavior model, demonstrate the effectiveness of the attention-based models. These new data-driven fire spread models exhibit promising modeling performances in both the next-step prediction (i.e., predicting fire progression from one timestep earlier) and recursive prediction (i.e., recursively predicting final fire perimeters from initial ignition points) of observed large wildfires in California, and they provide a foundation for further practical applications including short-term active fire spread prediction and long-term fire risk assessment. [ABSTRACT FROM AUTHOR]

Salinity management in estuarine systems is crucial for developing effective water-management strategies to maintain compliance and understand the impact of salt intrusion on water quality and availability. Understanding the temporal and spatial variations of salinity is a keystone of salinity-management practices. Process-based numerical models have been traditionally used to estimate the variations in salinity in estuarine environments. Advances in data-driven models (e.g., deep learning models) make them effective and efficient alternatives to process-based models. However, a discernible research gap exists in applying these advanced techniques to salinity modeling. The current study seeks to address this gap by exploring the innovative use of deep learning with data augmentation and transfer learning in salinity modeling, exemplified at 23 key salinity locations in the Sacramento–San Joaquin Delta which is the hub of the water-supply system of California. Historical, simulated (via a hydrodynamics and water quality model), and perturbed (to create a range of hydroclimatic and operational scenarios for data-augmentation purposes) flow, and salinity data are used to train a baseline multi-layer perceptron (MLP) and a deep learning Residual Long-Short-Term Memory (Res-LSTM) network. Four other deep learning models including LSTM, Residual Network (ResNet), Gated Recurrent Unit (GRU), and Residual GRU (Res-GRU) are also examined. Results indicate that models pre-trained using augmented data demonstrate improved performance over models trained from scratch using only historical data (e.g., median Nash–Sutcliffe efficiency increased from around 0.5 to above 0.9). Moreover, the five deep learning models further boost the salinity estimation performance in comparison with the baseline MLP model, though the performance of the latter is acceptable. The models trained using augmented data are then (a) used to develop a web-based Salinity Dashboard (Dashboard) tool that allows the users (including those with no machine learning background) to quickly screen multiple management scenarios by altering inputs and visualizing the resulting salinity simulations interactively, and (b) transferred and adapted to estimate observed salinity. The study shows that transfer learning results more accurately replicate the observations compared to their counterparts from models trained from scratch without knowledge learned and transferred from augmented data (e.g., median Nash–Sutcliffe efficiency increased from around 0.4 to above 0.9). Overall, the study illustrates that deep learning models, particularly when pre-trained using augmented data, are promising supplements to existing process-based models in estuarine salinity modeling, while the Dashboard enables user engagement with those pre-trained models to inform decision-making efficiently and effectively. [ABSTRACT FROM AUTHOR]

Shallow estuarine habitats are globally undergoing rapid changes due to climate change and anthropogenic influences, resulting in spatiotemporal shifts in distribution and habitat extent. Yet, scientists and managers do not always have rapidly available data to track habitat changes in real-time. In this study, we apply a novel and a state-of-the-art image segmentation machine learning technique (DeepLab) to two years of high-resolution drone-based imagery of a marine flowering plant species (eelgrass, a temperate seagrass). We apply the model to eelgrass (Zostera marina) meadows in the Morro Bay estuary, California, an estuary that has undergone large eelgrass declines and the subsequent recovery of seagrass meadows in the last decade. The model accurately classified eelgrass across a range of conditions and sizes from meadow-scale to small-scale patches that are less than a meter in size. The model recall, precision, and F1 scores were 0.954, 0.723, and 0.809, respectively, when using human-annotated training data and random assessment points. All our accuracy values were comparable to or demonstrated greater accuracy than other models for similar seagrass systems. This study demonstrates the potential for advanced image segmentation machine learning methods to accurately support the active monitoring and analysis of seagrass dynamics from drone-based images, a framework likely applicable to similar marine ecosystems globally, and one that can provide quantitative and accurate data for long-term management strategies that seek to protect these vital ecosystems. [ABSTRACT FROM AUTHOR]

• Propose a novel deep learning model based on Runge-Kutta recurrent neural network (RKRNN) with prior knowledge to realize structural system identification and seismic response prediction. • Formulate a partition training strategy to train the proposed neural network to improve the efficiency of training. • Utilize three numerical examples to valid the feasibility of RKRNN model including a linear three degrees of freedom system, a nonlinear single degree of freedom system with Bouc-Wen hysteresis and a simply supported bridge. • Site monitoring data from a bridge located in California has been used to further validate the proposed approach. In the seismic analysis of structural systems, dynamic response prediction is an essential problem and is significant in every stage during the structural life cycle. Conventionally, response analysis is carried out by numerical analysis. However, when the structural parameter is unknown, the establishment of a numerical model will be difficult. Enlightened by the Runge-Kutta (RK) numerical algorithm, this paper proposes a novel recurrent neural network named Runge-Kutta recurrent neural network (RKRNN) to realize the seismic response prediction. A partition training strategy is formulated to train the proposed neural network and to improve the efficiency of training. The proposed model can be trained by using a limited number of samples. Three numerical examples are utilized to validate the feasibility of RKRNN model including a linear three degrees of freedom (DOFs) system, a nonlinear single DOF system with Bouc-Wen hysteresis, and a numerical reinforced concrete bridge model. Additionally, the site monitoring data from a real-world bridge is utilized to further validate the proposed network. The results show that the proposed RKRNN model can effectively and efficiently predict the structural response under seismic load and exhibits robustness to noise, with good potential for applications in engineering practice. [ABSTRACT FROM AUTHOR]

Accurate forecasting of traffic flow is crucial for intelligent traffic control and guidance. It is very challenging to forecast the traffic flow due to the high non‐linearity, complexity and dynamicity of the data. Most existing forecasting methods focus on designing complicated graph neural network architectures to capture the spatio‐temporal features of traffic data with the help of predefined graphs. However, traffic data exhibit a strong spatial dependency, which means that there are often complex correlations between nodes in a road network topology graph. Moreover, the spatial correlations of the road network change over time. To capture the properties of the road network topology graph, a novel spatio‐temporal adaptive graph convolutional networks model (STAGCN) based on deep learning is proposed. Utilize adaptive graph generation block to capture the static and dynamic structures of the traffic road network respectively, and they are integrated to construct an adaptive road network topology graph. Then the spatio‐temporal features of the traffic data are captured using spatio‐temporal convolution blocks. Experiments were conducted on publicly available traffic datasets for freeway traffic in California, USA, and the results showed that the prediction accuracy of the STAGCN model outperformed multiple baseline methods. [ABSTRACT FROM AUTHOR]

Accurate estimation of fuels is essential for wildland fire simulations as well as decision-making related to land management. Numerous research efforts have leveraged remote sensing and machine learning for classifying land cover and mapping forest vegetation species. In most cases that focused on surface fuel mapping, the spatial scale of interest was smaller than a few hundred square kilometers; thus, many small-scale site-specific models had to be created to cover the landscape at the national scale. The present work aims to develop a large-scale surface fuel identification model using a custom deep learning framework that can ingest multimodal data. Specifically, we use deep learning to extract information from multispectral signatures, high-resolution imagery, and biophysical climate and terrain data in a way that facilitates their end-to-end training on labeled data. A multi-layer neural network is used with spectral and biophysical data, and a convolutional neural network backbone is used to extract the visual features from high-resolution imagery. A Monte Carlo dropout mechanism was also devised to create a stochastic ensemble of models that can capture classification uncertainties while boosting the prediction performance. To train the system as a proof-of-concept, fuel pseudo-labels were created by a random geospatial sampling of existing fuel maps across California. Application results on independent test sets showed promising fuel identification performance with an overall accuracy ranging from 55% to 75%, depending on the level of granularity of the included fuel types. As expected, including the rare—and possibly less consequential—fuel types reduced the accuracy. On the other hand, the addition of high-resolution imagery improved classification performance at all levels. [ABSTRACT FROM AUTHOR]

Water resources management in estuarine environments for water supply and environmental protection typically requires estimates of salinity for various flow and operational conditions. This study develops and applies two novel deep learning (DL) models, a residual long short-term memory (Res-LSTM) network, and a residual gated recurrent unit (Res-GRU) model, in estimating the spatial and temporal variations of salinity. Four other machine learning (ML) models, previously developed and reported, consisting of multi-layer perceptron (MLP), residual network (ResNet), LSTM, and GRU are utilized as the baseline models to benchmark the performance of the two novel models. All six models are applied at 23 study locations in the Sacramento–San Joaquin Delta (Delta), the hub of California's water supply system. Model input features include observed or calculated tidal stage (water level), flow and salinity at model upstream boundaries, salinity control gate operations, crop consumptive use, and pumping for the period of 2001–2019. Meanwhile, field observations of salinity at the study locations during the same period are also utilized for the development of the predictive use of the models. Results indicate that the proposed DL models generally outperform the baseline models in simulating and predicting salinity on both daily and hourly scales at the study locations. The absolute bias is generally less than 5%. The correlation coefficients and Nash–Sutcliffe efficiency values are close to 1. Particularly, Res-LSTM has slightly superior performance over Res-GRU. Moreover, the study investigates the overfitting issues of both the DL and baseline models. The investigation indicates that overfitting is not notable. Finally, the study compares the performance of Res-LSTM against that of an operational process-based salinity model. It is shown Res-LSTM outperforms the process-based model consistently across all study locations. Overall, the study demonstrates the feasibility of DL-based models in supplementing the existing operational models in providing accurate and real-time estimates of salinity to inform water management decision making. [ABSTRACT FROM AUTHOR]

Employing existing traffic data for the accurate classification of roads can provide significant references for the planning and construction of urban traffic infrastructure. This paper devises a novel deep learning framework for traffic station classification (DeepTSC) which classifies the traffic monitoring stations to in turn classify the road where these stations are located. The monitoring data of these traffic stations are interdependent with regards to both their spatial and temporal dimensions. We design a one-dimensional convolution layer for preliminary feature extraction and feature fusion. Following this, we employ a hybrid combination of the multilayer dilated convolution, long short-term memory network (LSTM) and attention mechanism, to extract the spatio-temporal features of the traffic data. Experiments on a public California freeway dataset show that the DeepTSC model's classification results outperform other existing state-of-the-art methods, with its accuracy being at least 1.4% higher than other models. [ABSTRACT FROM AUTHOR]

Systematically characterizing slip behaviours on active faults is key to unraveling the physics of tectonic faulting and the interplay between slow and fast earthquakes. Interferometric Synthetic Aperture Radar (InSAR), by enabling measurement of ground deformation at a global scale every few days, may hold the key to those interactions. However, atmospheric propagation delays often exceed ground deformation of interest despite state-of-the art processing, and thus InSAR analysis requires expert interpretation and a priori knowledge of fault systems, precluding global investigations of deformation dynamics. Here, we show that a deep auto-encoder architecture tailored to untangle ground deformation from noise in InSAR time series autonomously extracts deformation signals, without prior knowledge of a fault's location or slip behaviour. Applied to InSAR data over the North Anatolian Fault, our method reaches 2 mm detection, revealing a slow earthquake twice as extensive as previously recognized. We further explore the generalization of our approach to inflation/deflation-induced deformation, applying the same methodology to the geothermal field of Coso, California. A deep neural network is developed to automatically extract ground deformation from Interferometric Synthetic Aperture Radar time series. Applied to data over the North Anatolian Fault, the method can detect 2 mm deformation transients and reveals a slow earthquake twice as extensive as previously recognized. [ABSTRACT FROM AUTHOR]