The objective of this study is to predict road flooding risks based on topographic, hydrologic, and temporal precipitation features using machine learning models. Predictive flood monitoring of road network flooding status plays an essential role in
community hazard mitigation, preparedness, and response activities. Existing studies related to the estimation of road inundations either lack observed road inundation data for model validations or focus mainly on road inundation exposure assessment based on flood maps. This study addresses this limitation by using crowdsourced and fine-grained traffic data as an indicator of road inundation, and topographic, hydrologic, and temporal precipitation features as predictor variables. Two tree-based machine learning models (random forest and AdaBoost) were then tested and trained for predicting road inundations in the contexts of 2017 Hurricane Harvey and 2019 Tropical Storm Imelda in Harris County, Texas. The findings from Hurricane Harvey indicate that precipitation is the most important feature for predicting road inundation susceptibility, and that topographic features are more essential than hydrologic features for predicting road inundations in both storm cases. The random forest and AdaBoost models had relatively high AUC scores (0.860 and 0.810 for Harvey respectively and 0.790 and 0.720 for Imelda respectively) with the random forest model performing better in both cases. The random forest model showed stable performance for Harvey, while varying significantly for Imelda. This study advances the emerging field of smart flood resilience in terms of predictive flood risk mapping at the road level. For example, such models could help impacted communities and emergency management agencies develop better preparedness and response strategies with improved situational awareness of road inundation likelihood as an extreme weather event unfolds.
Atomic vibrations play a vital role in the functions of various physical, chemical, and biological materials. The vibrational properties and the specific heat of a bulk material are well described by the Debye theory, which successfully predicts the
quadratic $omega^{2}$ low-frequency scaling of the vibrational density of states (VDOS) in bulk solids from few fundamental assumptions. However, the corresponding relationships for nanoconfined materials with fewer degrees of freedom have been far less well explored. In this work, using inelastic neutron scattering, we characterize the VDOS of amorphous ice confined to a thickness of $approx 1$ nm inside graphene oxide membranes and we observe a crossover from the Debye $omega^2$ scaling to a novel and anomalous $omega^3$ behaviour upon reducing the confinement size $L$. Additionally, using molecular dynamics simulations, we not only confirm the experimental findings but also prove that such a novel scaling of the VDOS appears in both crystalline and amorphous solids under slab-confinement. Finally, we theoretically demonstrate that this low-frequency $omega^3$ law results from the geometric constraints on the momentum phase space induced by confinement along one spatial direction. This new physical phenomenon, revealed by combining theoretical, experimental and simulations results, is relevant to a myriad of systems both in synthetic and biological contexts and it could impact various technological applications for systems under confinement such as nano-devices or thin films.
The objective of this research is to explore the temporal importance of community-scale human activity features for rapid assessment of flood impacts. Ultimate flood impact data, such as flood inundation maps and insurance claims, becomes available o
nly weeks and months after the floods have receded. Crisis response managers, however, need near-real-time data to prioritize emergency response. This time lag creates a need for rapid flood impact assessment. Some recent studies have shown promising results for using human activity fluctuations as indicators of flood impacts. Existing studies, however, used mainly a single community-scale activity feature for the estimation of flood impacts and have not investigated their temporal importance for indicating flood impacts. Hence, in this study, we examined the importance of heterogeneous human activity features in different flood event stages. Using four community-scale big data categories we derived ten features related to the variations in human activity and evaluated their temporal importance for rapid assessment of flood impacts. Using multiple random forest models, we examined the temporal importance of each feature in indicating the extent of flood impacts in the context of the 2017 Hurricane Harvey in Harris County, Texas. Our findings reveal that 1) fluctuations in human activity index and percentage of congested roads are the most important indicators for rapid flood impact assessment during response and recovery stages; 2) variations in credit card transactions assumed a middle ranking; and 3) patterns of geolocated social media posts (Twitter) were of low importance across flood stages. The results of this research could rapidly forge a multi-tool enabling crisis managers to identify hotspots with severe flood impacts at various stages then to plan and prioritize effective response strategies.
The objective of this study is to predict the near-future flooding status of road segments based on their own and adjacent road segments current status through the use of deep learning framework on fine-grained traffic data. Predictive flood monitori
ng for situational awareness of road network status plays a critical role to support crisis response activities such as evaluation of the loss of access to hospitals and shelters. Existing studies related to near-future prediction of road network flooding status at road segment level are missing. Using fine-grained traffic speed data related to road sections, this study designed and implemented three spatio-temporal graph convolutional network (STGCN) models to predict road network status during flood events at the road segment level in the context of the 2017 Hurricane Harvey in Harris County (Texas, USA). Model 1 consists of two spatio-temporal blocks considering the adjacency and distance between road segments, while Model 2 contains an additional elevation block to account for elevation difference between road segments. Model 3 includes three blocks for considering the adjacency and the product of distance and elevation difference between road segments. The analysis tested the STGCN models and evaluated their prediction performance. Our results indicated that Model 1 and Model 2 have reliable and accurate performance for predicting road network flooding status in near future (e.g., 2-4 hours) with model precision and recall values larger than 98% and 96%, respectively. With reliable road network status predictions in floods, the proposed model can benefit affected communities to avoid flooded roads and the emergency management agencies to implement evacuation and relief resource delivery plans.
In this study, an order by disorder mechanism has been proposed in a two-dimensional PXP model, where the extensive degeneracy of the classical ground-state manifold is due to strict occupation constraints instead of geometrical frustrations. By perf
orming an unbias large-scale quantum monte carlos simulation, we find that local quantum fluctuations, which usually work against long-range ordering, lift the macroscopic classical degeneracy and give rise to a compressible ground state with charge-density-wave long-range order. A simple trial wavefunction has been proposed to capture the essence of the ground-state of the two-dimensional PXP model. The finite temperature properties of this model have also been studied, and we find a thermal phase transition with an universality class of two-dimensional Ising model.
Moire superlattices in van der Waals heterostructures are gaining increasing attention because they offer new opportunities to tailor and explore unique electronic phenomena when stacking 2D materials with small twist angles. Here, we reveal local su
rface potentials associated with stacking domains in twisted double bilayer graphene (TDBG) moire superlattices. Using a combination of both lateral Piezoresponse Force Microscopy (LPFM) and Scanning Kelvin Probe Microscopy (SKPM), we distinguish between Bernal (ABAB) and rhombohedral (ABCA) stacked graphene and directly correlate these stacking configurations with local surface potential. We find that the surface potential of the ABCA domains is ~15 mV higher (smaller work function) than that of the ABAB domains. First-principles calculations based on density functional theory further show that the different work functions between ABCA and ABAB domains arise from the stacking dependent electronic structure. We show that, while the moire superlattice visualized by LPFM can change with time, imaging the surface potential distribution via SKPM appears more stable, enabling the mapping of ABAB and ABCA domains without tip-sample contact-induced effects. Our results provide a new means to visualize and probe local domain stacking in moire superlattices along with its impact on electronic properties.
We introduce the matrix-based Renyis $alpha$-order entropy functional to parameterize Tishby et al. information bottleneck (IB) principle with a neural network. We term our methodology Deep Deterministic Information Bottleneck (DIB), as it avoids var
iational inference and distribution assumption. We show that deep neural networks trained with DIB outperform the variational objective counterpart and those that are trained with other forms of regularization, in terms of generalization performance and robustness to adversarial attack.Code available at https://github.com/yuxi120407/DIB
Measuring the dependence of data plays a central role in statistics and machine learning. In this work, we summarize and generalize the main idea of existing information-theoretic dependence measures into a higher-level perspective by the Shearers in
equality. Based on our generalization, we then propose two measures, namely the matrix-based normalized total correlation ($T_alpha^*$) and the matrix-based normalized dual total correlation ($D_alpha^*$), to quantify the dependence of multiple variables in arbitrary dimensional space, without explicit estimation of the underlying data distributions. We show that our measures are differentiable and statistically more powerful than prevalent ones. We also show the impact of our measures in four different machine learning problems, namely the gene regulatory network inference, the robust machine learning under covariate shift and non-Gaussian noises, the subspace outlier detection, and the understanding of the learning dynamics of convolutional neural networks (CNNs), to demonstrate their utilities, advantages, as well as implications to those problems. Code of our dependence measure is available at: https://bit.ly/AAAI-dependence
The objective of this study is to examine spatial patterns of impacts and recovery of communities based on variances in credit card transactions. Such variances could capture the collective effects of household impacts, disrupted accesses, and busine
ss closures, and thus provide an integrative measure for examining disaster impacts and community recovery in disasters. Existing studies depend mainly on survey and sociodemographic data for disaster impacts and recovery effort evaluations, although such data has limitations, including large data collection efforts and delayed timeliness results. In addition, there are very few studies have concentrated on spatial patterns and disparities of disaster impacts and short-term recovery of communities, although such investigation can enhance situational awareness during disasters and support the identification of disparate spatial patterns of disaster impacts and recovery in the impacted regions. This study examines credit card transaction data Harris County (Texas, USA) during Hurricane Harvey in 2017 to explore spatial patterns of disaster impacts and recovery during from the perspective of community residents and businesses at ZIP code and county scales, respectively, and to further investigate their spatial disparities across ZIP codes. The results indicate that individuals in ZIP codes with populations of higher income experienced more severe disaster impact and recovered more quickly than those located in lower-income ZIP codes for most business sectors. Our findings not only enhance the understanding of spatial patterns and disparities in disaster impacts and recovery for better community resilience assessment, but also could benefit emergency managers, city planners, and public officials in harnessing population activity data, using credit card transactions as a proxy for activity, to improve situational awareness and resource allocation.
Optical and optoelectronic approaches of performing matrix-vector multiplication (MVM) operations have shown the great promise of accelerating machine learning (ML) algorithms with unprecedented performance. The incorporation of nanomaterials into th
e system can further improve the performance thanks to their extraordinary properties, but the non-uniformity and variation of nanostructures in the macroscopic scale pose severe limitations for large-scale hardware deployment. Here, we report a new optoelectronic architecture consisting of spatial light modulators and photodetector arrays made from graphene to perform MVM. The ultrahigh carrier mobility of graphene, nearly-zero-power-consumption electro-optic control, and extreme parallelism suggest ultrahigh data throughput and ultralow-power consumption. Moreover, we develop a methodology of performing accurate calculations with imperfect components, laying the foundation for scalable systems. Finally, we perform a few representative ML algorithms, including singular value decomposition, support vector machine, and deep neural networks, to show the versatility and generality of our platform.
