Hurricanes cause significant property loss every year. A substantial part of that loss is due to trees destroyed by wind. The destroyed trees block roads and produce a large amount of debris that needs to be cleaned after the hurricane. This paper presents a modular framework to assess the consequence of tree failure due to high winds, focusing on the road closure probability and the debris estimation. Key modules of the proposed framework include Convolutional Neural Networks (CNNs) for tree recognition, tree failure assessment, and tree failure consequences. A significant difference between the proposed framework and other assessment models is that the proposed framework can estimate the road closure probability and the amount of debris before the hurricane, which may help authorities to prevent the potential loss by acting before it happens. CNNs for tree recognition module uses satellite imagery as an input and is composed of two networks. Network-I recognizes the tree locations and Network-II classifies the tree species. The tree failure assessment module uses the selected tree images to approximate the tree parameters and calculate the fragility functions for all the trees that can have an impact in the affected area. The last module estimates the consequence of tree failure such as the fragility of a road, a probability of road closure, and the amount of tree debris produced by the hurricane. As an illustrative example, a region in the City of Tallahassee, Florida has been investigated. The results derived by the proposed approach were validated based on the data obtained after Hurricane Hermine. The proposed approach can help city and state authorities to better plan for the adverse consequences of tree failures caused by hurricane winds.

Florida has distinctively different regions, in terms of population density and hurricane risk. The objective of the article is to understand the characteristics of hurricane loss in different regions of Florida. Two different structures with different components capacities were taken into consideration. The first structure does not meet the current building standards in terms of the high wind performance (‘weak’) and the second structure is built from hurricane proved materials approved by Florida Building Code (‘strong’). In order to bring a new insight into the regional differences in hurricane loss, Exceedance Probability of Loss (EPL) is introduced. The EPL is the probability that the total annual loss will exceed a certain level. The results of this research identified the most vulnerable parts of Florida in terms of the EPL. It has been shown how to combine results of analyses conducted with percentage loss and dollar loss to arrive at a fuller understanding in regional analyses. Finally, the paper also discussed the importance of roof-to-wall connections in preventing the roof uplift.

This paper uses a Bayesian capacity model to approximate the loss of low-rise buildings from hurricane damage. Many previous vulnerability models utilized components capacities using an analytical estimation composed with a statistical distribution of an error. In this paper, a Bayesian capacity model that combines physics-based analytical results with experimental results was used in the vulnerability estimation. The inclusion of analytical estimations can help to produce a more accurate fragility function when the amount of available data is limited. The Bayesian approach allows a proper treatment of model uncertainties and inherent uncertainties. The proposed approach provides a range of the fragility functions and their confidence intervals, which then is translated into upper and lower bounds in the vulnerability estimation. To demonstrate the advantages of the proposed approach, the vulnerability of a wood-frame house was estimated using Bayesian capacity models of roof structural components and compared to a conventional vulnerability model. It was shown that not accounting for model uncertainty may result in an underestimation of the potential loss compared to the proposed Bayesian approach.