Hurricane risk analysis: A review on the physically-based approach

This paper reviews recent studies that take a physically-based approach to better assess and manage hurricane risk. Such a methodology includes three components: modeling the storm climatology (which defines TC risk in terms of the upper tail of the storm statistics); modeling landfalling hazards; and characterizing damage and losses. not rely on the limited historical TC data but generates large samples of synthetic storms that are in statistical agreement with observations (Emanuel et al. 2006), and compares well with other methods used to study the effects of climate change on TCs (Emanuel et al. 2008 and 2010; Bender et al. 2010; Knutson et al. 2010). This approach has been used to investigate TC wind and surge risk (e.g. Lin et al. 2010 and 2012; Klima et al. 2011; Mendelsohn et al. 2012). Here we briefly summarize the Emanuel et al. method for generating synthetic storms. Storm genesis points are randomly selected from a distribution constructed from the Best-Track historical dataset or generated by a random seeding technique. Once initiated, storm displacements are calculated using the ‘beta and advection model’ in which 850 and 250 mb environmental steering flows vary randomly but in accordance with the monthly mean, variance and covariances of the reanalysis data or the climate model prediction. Along each simulated track, the Coupled Hurricane Intensity Prediction System (CHIPS), a deterministic ocean-atmospheric coupled model, is used to simulate the intensity evolution of the storm (Emanuel et al. 2004). TC risk for a certain area is estimated by examining the frequency and characteristics of storms passing by the area (e.g., crossing a boundary of the area or passing within a certain radius of the center of the area). The CHIPS model also estimates the storm radius of maximum wind, conditional on the storm outer radius (where the storm wind vanishes). The storm outer radius was found to obey a lognormal distribution (Chavas and Emanuel 2010). Theoretically, the storm outer radius scales with the storm potential intensity (Emanuel 1986); however, the scaling relationship remains to be evaluated against observations; at the same time, further investigation is needed to determine how storm outer radii will change with climate. Storm size (as described by the storm radius of maximum wind, the outer radius, and other related parameters) is an important factor, in addition to storm intensity and track, in determining TC hazards, as amply demonstrated by Hurricanes Katrina, Irene, and Sandy. 3 TC HAZARD MODELING TCs induce multiple hazards during and after landfall, including strong winds, heavy rainfall, and storm surge. TC wind and rainfall may be simulated using high-resolution dynamic models, such as the Weather Research and Forecasting (WRF; Skamarock et al. 2005) model. For example, Lin et al. (2010b) carried out a case study for Hurricane Isabel (2003), applying the WRF model to examine storm properties and wind and rainfall hazards down to 1km resolution at landfall. They further coupled the WRF model with a hydrodynamic model to simulate the storm surge in Chesapeake Bay. This approach may be applied to studying the impact of climate change on TC hazards; in particular, high-resolution dynamic models can be used to downscale relatively lower-resolution TC simulations, such as Knutson et al.’s (2012 and 2013) projection of storms in different climates, to local scales. However, this approach is, at present, computationally too expensive to be applied directly to risk assessment, which should involve very large numbers (~10) of simulations to cover many possible scenarios. Consequently, computationally much more efficient parametric models are often used in risk analysis. Parametric wind and pressure models can also be applied to storm surge analysis. Such parametric modeling and analysis can be readily coupled with TC risk models (see above) to estimate TC hazard risk. In this section, we review parametric wind modeling and storm surge modeling as well as statistical methods to estimate TC hazard risk from physical model results. A physically-based parametric rainfall model, which depends on the wind model, is currently under development by Emanuel.