Toward a refined estimation of typhoon wind hazards: Parametric modeling and upstream terrain effects

Highlights

• A dataset of two key typhoon wind field parameters, i.e. $R_{max,s}$ and $B_s$ in Western Pacific Ocean was developed.

• The statistically recursive models of $R_{max,s}$ and $B_s$ were proposed.

• An improved algorithm accounting for the directional upstream terrain effects on wind speed was present.

• The wind hazard footprints of 184 observed landed or offshore typhoon-scale storms were generated.

Abstract

Parametric and stochastic typhoon model enable a rapid estimation of wind hazards in typhoon-prone regions. It always needs a large amount of historical track information to develop the track, intensity as well as wind field parameters. This study describes a technique for estimating two commonly used typhoon wind field parameters, i.e. $R_{max,s}$ and $B_s$ using the observed wind information from the best track dataset coupled with a semi-analytical wind field model. At each timestep of every typhoon event, the radial wind speed profile is well reproduced with an optimal pair of $R_{max,s}$ and $B_s$. The correlation analyses of $R_{max,s}$ and $B_s$ with other parameters are conducted. The $R_{max,s}$ and $B_s$ at different timesteps allows the development of recursive models accounting for their autocorrelations between adjacent timesteps. Linearly weighted progressive formulas of $R_{max,s}$ and $B_s$ using all data extracted in the Western Pacific domain are developed. This idea is similar to the track and intensity models during stochastic typhoon simulations, which provides a forward step towards the more rational estimation of typhoon hazards. Moreover, the typhoon-event-specific $R_{max,s}$ and $B_s$ enables the reconstruction of historical wind hazards. By introducing the underlying terrain effects on wind speeds in terms of a directional roughness length and a topographic speed-up factor, 184 observed landed or offshore typhoon-scale storms along the China coastline from 1977 to 2015 are investigated. A dataset regarding the wind hazard footprints for over-water, roughness only and roughness and topography combined conditions of these 184 storms is developed to facilitate risk assessment and disaster mitigation during typhoon events.

Introduction

Typhoon or hurricane (typhoon hereafter is a general representation of tropical cyclone unless otherwise stated) is a large-scale air rotating system around a low atmospheric pressure center, frequently causing devastating economic loss and human casualties along coastal regions due to violent winds, heavy rainfall, massive storm surges, flash flooding or even landslides in mountainous areas. The Western Pacific Basin is the most active typhoon basin on earth, accounting for almost one-third of global annual storms. Seven or eight typhoons per year make landfall over the coastal region of China, where is characterized by high population densities and highly developed cities. It was estimated that averaged 472 people lost their lives and annual direct economic loss reached 28.7 billion RMB as a result of landfall typhoons from year 1983–2016 in China (Zhang et al., 2009). The losses are expected to continuously increase because of growing population and increasing wealth in coastal regions as well as the potential rise of typhoon frequency and intensity due to climate change. The investigation of typhoon wind characteristics and prediction of typhoon-induced hazards are of great importance to facilitate the disaster prevention and mitigation (Fang et al., 2019, 2020).

The quantification of typhoon boundary layer with the depth about 2–3 ​km, within which we live and carry out most human activities, has received intensive attention in past several decades (Batts et al., 1980; Meng et al., 1995; Vickery et al., 2000a, Vickery et al., 2000b; Kepert, 2010; Snaiki and Wu, 2017; Fang et al., 2018a) for the uses of engineering applications and wind hazard estimations. Parametric typhoon wind field model was commonly adopted for its high efficiency on Monte Carlo simulations as well as its continuous improvements with the help of the more measurement data. Recent years, the ever-increasing observation data have enabled a further investigation on typhoon inner structures. Taking the advantages of flight-level aircraft and dropsondes measurements in Atlantic Basin, a series of pioneering studies have been conducted to examine the characteristics of two typical typhoon field parameters, say the radius to maximum wind speed ($R_{max}$) and radial pressure profile shape parameter ($B$), and model them with some statistically-based equations for the convenience of stochastic simulations (Powell et al., 2005; Vickery and Wadhera, 2008). Recently, some parameter models have also been successively developed in Western Pacific region using observation data (Xiao et al., 2011; Zhao et al., 2013; Fang et al., 2018b).

However, several issues remain to be discussed about these two parameters with respect to their height-variation, region-dependent and time duration characteristics. As discussed by Willoughby and Rahn (2004), it showed that the mean value of $R_{max}$ increases slightly with height while $B$ witnesses a 45% increase from the altitude about 750 ​m–2500 ​m based on a flight-level database. Holland et al. (2010) also tried to revise the pressure-wind model by addressing the differences between surface and gradient layers. In addition, these typhoon field parameters are usually region-dependent due to the difference of atmospheric circulation features, suggesting the cross-adoption of these parameter models could result in some unreasonable predictions. Furthermore, the agency-specified wind speed averaging period varies considerably (from 1 ​min to 10 ​min), resulting in the difference of central pressure estimation based on Dvorak method (Dvorak, 1984; Velden et al., 2006). This could be extended to the misunderstanding of $R_{max}$ and $B$ during their extractions and applications. Another issue is that the statistical models of $R_{max}$ and $B$ were always formulated as the function of central pressure deficit, typhoon center latitude and sea surface temperature $T_s$. The autocorrelations between different timesteps are indirectly propagated from the central pressure deficit, and sea surface temperature $T_s$ during empirical full track simulations. This could result in the storm structure and size fluctuate notably with timesteps, which is inconsistent with the real cases.

To extract the parameters of $R_{max}$ and $B$, a parametric typhoon pressure or wind field model should be introduced. However, most parametric wind field models are simplified from Navier-Stokes equations, i.e. several nonlinear terms and non-symmetric characteristics are customarily eliminated. If the Holland parametric pressure model described by $R_{max}$ and $B$ is derived from real pressure observations, the pressure field would be well reconstructed. But it could lead to unreal wind field due to the use of simplified model solutions. Alternatively, if $R_{max}$ and $B$ are extracted from the fitting results of real winds, the modeled wind field is as close to the reality as possible regardless of whether the pressure field is real or unreal. This can be achieved using the archived wind information in some best track dataset, such as HURDAT2 in Atlantic Basin and RSMC Best Track Data in Northwestern Pacific Ocean provided by Japan Meteorological Agency. It also allows the consideration of autocorrelations $R_{max}$ and $B$ between different timesteps to better conduct the stochastic simulations of wind hazard. Moreover, the evolutions of wind speed for each historical typhoon event can be reconstructed to facilitate the typhoon hazard assessment and mitigation.

Conventionally, to reproduce the over-land wind snapshots or footprints of typhoon storms, the underlying exposure was assumed as the open-flat terrain. However, local terrain roughness and topographic features as well as surrounding obstacles would determine the development of a boundary layer and evolution of turbulence. In reality, a sudden change of elevation or topography would have an obvious impact on surface wind speed over a very short distance (Miller et al., 2013), especially the speeds near the crests of ridges and hills, which show marked increases when compared with the wind speed measured at same height above the flat terrain. Some studies (Lemelin et al., 1988; Weng et al., 2000) found that the wind speed at top of the hill could even double the speed that over flat terrain due to topographic effects, which represents a structure on top of hill would experience an increase of 400% in the wind load than that in flat area. Accordingly, quantification of directional roughness and topographic effects is essentially important for typhoon wind hazards assessments. Some pioneer studies (Jackson and Hunt, 1975; Taylor et al., 1983; Hunt et al., 1988) have well developed the theory of boundary layer flow over low-slope topography. These methods provide a good estimation of topography effects for low slope hills but require massive computation resource if they are applied to a large area.

In this study, $R_{max}$ and $B$ at surface level are optimally fitted with a high-resolving typhoon boundary layer wind field model using the JMA best track dataset. The correlations between multiple typhoon field parameters are investigated before the development of recursive models for $R_{max}$ and $B$ accounting for the autocorrelations with previous timesteps. They are used to reconstruct the wind hazards of historical typhoon events. The upstream roughness and topographic effects for sites of interest are quantitatively estimated with a directional equivalent roughness length and a topographic speed-up factor.