Reanalysing strong-convective wind damage paths using high-resolution aerial images

Abstract

High-resolution aerial imagery may provide very detailed information about strong-convective wind events, which can be very useful to enhance and make more robust severe weather databases. Through an aerial analysis, unknown affected areas can be unveiled, large damaged zones or complex terrain events can be mapped, and in situ damage surveys can be completed. Comparing aerial orthophotographs or high-resolution satellite imagery taken before and after the event, damage to forest associated with EF1 + intensities rated with the Enhanced Fujita scale is easily observable. It is also possible to detect repaired parts on some types of roofs and buildings due to the occurrence of weak events and major damage on human-made structures in case of significant episodes. To illustrate the added value of using orthophotographs in forensic damage survey analysis, the 2 November 2008 strong-convective wind event in Catalonia (NE Iberian Peninsula) is revisited. Confirming previous studies, two different damage swaths, associated with the passage of two supercells during the early morning, are clearly identified in high-resolution aerial images. The additional information provided by orthophotograph analysis allowed extending the two damage paths and increasing the damage intensity of the event. It is concluded that both damage tracks are related to tornadoes, which are rated as an EF2 and an EF3, the second being the strongest tornado ever registered in Catalonia, one of the most tornado-prone regions of southern Europe.

Appendix: Wind intensity versus damage path width and length

Given a Weibull distribution (Brooks 2004):

$$f\left( x \right) = \left( {\frac{\alpha }{\beta }} \right)\left( {\frac{x}{\beta }} \right)^{\alpha - 1} \exp \left[ { - \left( {\frac{x}{\beta }} \right)^{\alpha } } \right],$$

(2)

the cumulative distribution function is:

$$F\left( x \right) = 1 - \exp \left[ { - \left( {\frac{x}{\beta }} \right)^{\alpha } } \right].$$

(3)

Considering the damage path length and width as independent variables (Shikhov and Chernokulsky 2018):

$$P_{LW} = P_{L} + P_{W} - P_{L} P_{W}$$

(4)

where $P_{LW}$ is the probability of exceeding an F-scale intensity given the length (L) and width (W) of the damage path, $P_{L}$ is the probability of exceeding an F-scale intensity given the length (L) (i.e. cumulative distribution function for L, $F\left( {x = L} \right)$) and $P_{W}$ is the probability of exceeding an F-scale intensity given the width (W) (i.e. cumulative distribution function for W, $F\left( {x = W} \right)$).

Replacing $P_{L}$ by the cumulative distribution function for the length and $P_{W}$ by the cumulative distribution function for the length in Eq. 4:

$$P_{LW} = 1 - \exp \left[ { - \left( {\frac{L}{{\beta_{L} }}} \right)^{{\alpha_{L} }} } \right] + 1 - \exp \left[ { - \left( {\frac{W}{{\beta_{W} }}} \right)^{{\alpha_{W} }} } \right] - \left\{ {1 - \exp \left[ { - \left( {\frac{L}{{\beta_{L} }}} \right)^{{\alpha_{L} }} } \right]} \right\}\left\{ {1 - \exp \left[ { - \left( {\frac{W}{{\beta_{W} }}} \right)^{{\alpha_{W} }} } \right]} \right\}$$

(5)

$$P_{LW} = 1 - \exp \left[ { - \left( {\frac{L}{{\beta_{L} }}} \right)^{{\alpha_{L} }} - \left( {\frac{W}{{\beta_{W} }}} \right)^{{\alpha_{W} }} } \right]$$

(6)

and, arranging the equation to present it as function (W) depending on length (L), it is:

$$W\left( L \right) = \beta_{W} \left[ { - \left( {\frac{L}{{\beta_{L} }}} \right)^{{\alpha_{L} }} - \ln \left( {1 - P_{LW} } \right)} \right]^{{1/\alpha_{W} }}.$$

(7)

Then, it is possible to make the plot presented in the manuscript (Fig. 5), using $\alpha_{L}$, $\beta_{L}$ (for length), and $\alpha_{W}$, $\beta_{W}$ (for width) from Brooks (2004) for each F-scale degree, and being $P_{LW} = 0.70$ for discontinuous lines and $P_{LW} = 0.90$ for continuous lines.