Effects of damage initiation points of depth-damage function on flood risk assessment.

The flood depth in a structure is a key factor in flood loss models, influencing the estimation of building and contents losses, as well as overall flood risk. Recent studies have emphasized the importance of determining the damage initiation point (DIP) of depth-damage functions, where the flood damage is assumed to initiate with respect to the first-floor height of the building. Here we investigate the effects of DIP selection on the flood risk assessment of buildings located in Special Flood Hazard Areas. We characterize flood using the Gumbel extreme value distribution's location (µ) and scale (α) parameters. Results reveal that average annual flood loss (AAL) values do not depend on µ, but instead follow an exponential decay pattern with α when damage initiates below the first-floor height of a building (i.e., negative DIP). A linear increasing pattern of the AAL with α is achieved by changing the DIP to the first-floor height (i.e., DIP = 0). The study also demonstrates that negative DIPs have larger associated AAL, thus contributing substantially to the overall AAL, compared to positive DIPs. The study underscores the significance of proper DIP selection in flood risk assessment.

A data-driven spatial approach to characterize the flood hazard.

Model output of localized flood grids are useful in characterizing flood hazards for properties located in the Special Flood Hazard Area (SFHA-areas expected to experience a 1% or greater annual chance of flooding). However, due to the unavailability of higher return-period [i.e., recurrence interval, or the reciprocal of the annual exceedance probability (AEP)] flood grids, the flood risk of properties located outside the SFHA cannot be quantified. Here, we present a method to estimate flood hazards that are located both inside and outside the SFHA using existing AEP surfaces. Flood hazards are characterized by the Gumbel extreme value distribution to project extreme flood event elevations for which an entire area is assumed to be submerged. Spatial interpolation techniques impute flood elevation values and are used to estimate flood hazards for areas outside the SFHA. The proposed method has the potential to improve the assessment of flood risk for properties located both inside and outside the SFHA and therefore to improve the decision-making process regarding flood insurance purchases, mitigation strategies, and long-term planning for enhanced resilience to one of the world's most ubiquitous natural hazards.

Flood risk assessment for residences at the neighborhood scale by owner/occupant type and first-floor height.

Evaluating flood risk is an essential component of understanding and increasing community resilience. A robust approach for quantifying flood risk in terms of average annual loss (AAL) in dollars across multiple homes is needed to provide valuable information for stakeholder decision-making. This research develops a computational framework to evaluate AAL at the neighborhood level by owner/occupant type (i.e., homeowner, landlord, and tenant) for increasing first-floor height (FFH). The AAL values were calculated here by numerically integrating loss-exceedance probability distributions to represent economic annual flood risk to the building, contents, and use. A simple case study for a census block in Jefferson Parish, Louisiana, revealed that homeowners bear a mean AAL of $4,390 at the 100-year flood elevation (E 100), compared with $2,960, and $1,590 for landlords and tenants, respectively, because the homeowner incurs losses to building, contents, and use, rather than only two of the three, as for the landlord and tenant. The results of this case study showed that increasing FFH reduces AAL proportionately for each owner/occupant type, and that two feet of additional elevation above E 100 may provide the most economically advantageous benefit. The modeled results suggested that Hazus Multi-Hazard (Hazus-MH) output underestimates the AAL by 11% for building and 15% for contents. Application of this technique while partitioning the owner/occupant types will improve planning for improved resilience and assessment of impacts attributable to the costly flood hazard.

Modeling Focused-Ultrasound Response for Non-Invasive Treatment Using Machine Learning.

The interactions between body tissues and a focused ultrasound beam can be evaluated using various numerical models. Among these, the Rayleigh-Sommerfeld and angular spectrum methods are considered to be the most effective in terms of accuracy. However, they are computationally expensive, which is one of the underlying issues of most computational models. Typically, evaluations using these models require a significant amount of time (hours to days) if realistic scenarios such as tissue inhomogeneity or non-linearity are considered. This study aims to address this issue by developing a rapid estimation model for ultrasound therapy using a machine learning algorithm. Several machine learning models were trained on a very-large dataset (19,227 simulations), and the performance of these models were evaluated with metrics such as Root Mean Squared Error (RMSE), R-squared (R2), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). The resulted random forest provides superior accuracy with an R2 value of 0.997, an RMSE of 0.0123, an AIC of -82.56, and a BIC of -81.65 on an external test dataset. The results indicate the efficacy of the random forest-based model for the focused ultrasound response, and practical adoption of this approach will improve the therapeutic planning process by minimizing simulation time.

A deep learning based traffic crash severity prediction framework.

Highway work zones are most vulnerable roadway segments for congestion and traffic collisions. Hence, providing accurate and timely prediction of the severity of traffic collisions at work zones is vital to reduce the response time for emergency units (e.g., medical aid), accordingly improve traffic safety and reduce congestion. In predicting the severity of traffic collisions, previous studies used different statistical and machine learning models with accuracy as the main evaluating factor. However, the performance of these models was generally not good, especially on fatal and injury crashes. Also, looking into the prediction accuracy only is misleading. This paper aims to propose a novel deep learning-based approach with a customized f1-loss function to predict the severity of traffic crashes. Underlying this objective is to compare the results of deep learning models with machine learning model considering two performance indicators, namely precision, and recall. The data used in the analysis include a sample of traffic crashes that occurred at work zones in Louisiana from 2014 to 2018. This dataset includes valuable information (features) related to road, vehicle, and human factors affecting the occurrence and severity of those crashes. The proposed methodology is based on transforming these features/variables into images. Image transformation is conducted using a nonlinear dimensionality reduction technique t-SNE and convex hull algorithm. A CNN based deep learning algorithm with a customized loss function was used to directly optimize the model for precision and recall. The results showed improved performance in predicting the crash severity of fatal and injury crashes using the deep learning approach, which can help to improve traffic safety as well as traffic congestion at work zones and possibly other roadways segments.