Wednesday, 6 April 2016

Remote damage survey indicates tornado was culprit of Wyndham Vale damage on 18 March 2016

After this event occurred it took only a few hours for the Bureau of Meteorology to dismiss the possibility that a tornado had occurred (Irving, 2016C). I started this remote damage survey expecting the event not to be a tornado, possibly a microburst, but most likely I expected this survey to be inconclusive. 

There was one potentially credible witness who said, "I've only heard that noise once before and it sounded like a mini-tornado" (ABC, 2016). Identifying a tornado by only its sound because you've heard one before maybe possible, but unlikely to suffice as good evidence.

Radar data available publicly in Australia presents a significant problem when it comes to analysis to identify features related to tornadoes. This is because it is all composite radar images, not the base reflectivity scans typically needed. The Doppler radar wind at the time has subtle, but identifiable indications of mesocyclonic activity associated with the northern end of a bowing line segment of storms visible on the rain rate radar.

The remote damage survey has used open source data mining of social media; searching websites including Google, Twitter and Facebook for photos and videos of the damage. The potential of crowd sourcing photographs (Baker & Foulser-Piggot, 2015; Kim, Pant & Yamashita, 2015) and social media data mining (Kryvasheyeu et al. 2015) as post-disaster damage assessment vehicles has been well recognized. After the Moore, Oklahoma tornado of 2013 passive and active data mining of social media was used to identify damage locations within the bounded storm areas using geo-located tweets in real-time (Graettinger et al. 2014, p. 29). The images mined from Twitter after the Moore tornado proved useful in establishing how much cleanup had taken place before personnel from the damage survey team had arrived to make their assessment.

Once images of damage were sourced, those that had associated location information in captions, articles or associated posts were utilized first. Locations were searched for on Google Maps, satellite/aerial imagery enabled specific identifiable features in pictures to be located from above. These locations were then checked via Google Street View for accuracy, just as done during the Moore 2013 damage survey (Graettinger et al. 2014, p. 27). Pictures that lacked location information were then located by visually searching the areas in proximity to the know location of damage.

Once complete 13 locations were identified as having sustained damage at Manor Lakes estate in Wyndham Vale. When mapped, the damage formed a line just over 500 meters long and little more than a house wide (mostly less than 25 metres, max width 40 metres). When a straight line is plotted from the location of damage furthest west to the location of damage furthest east, all the observed damage has occurred within 20 metres of that line (map markers placed in the centre of damaged structures).

Not only did the damage form a narrow defined path, it also presented wild variations in wind direction (completely reversing direction in less than 10 metres at one point). The final evidence that points clearly at a tornado are three examples of items being picked up and carried unknown distances away (I did not find reports they were found). These items included a trampoline and air-conditioning units from two separate houses. The removal of air-conditioning units where they have been carried away is indicative of a tornado, not straight line winds or a microburst.

Using the Enhanced Fujita (EF) Scale to estimate wind speeds, this tornado was an EF0 with winds 105-137 kph. Almost all (12 out of 13) of the damage indicators were one and two family residences (FR12 DI2), of these 6 had sustained the threshold of visible damage (DOD1 Expected 118 kph, Lower bound 98 kph, Upper bound 141 kph). The other 6 had sustained a degree of damage that included the loss of roof covering material (<20%) and/or gutters (DOD2 Expected 126 kph, Lower bound 101 kph, Upper bound 155 kph). The last damage indicator was an adolescent eucalyptus tree (hardwood) having all its large branches (25-76 mm [1-3 inches] in diameter) broken (TH DI27 DOD2 Expected 118 kph, Lower bound 98 kph, Upper bound 141 kph). If considered akin to the hardwood trees categorized in the EF Scale, this can be a usable damage indicator.

The new construction of all these buildings gives confidence in ascribing the expected wind speeds for the degree of damage done (this assumes they met building codes). The proximity of damaged structures to one another provides excellent contextual and numerical confirmation for this rating.

Although this study is not comprehensive, focusing heavily on the damage evidence, it does makes the case that further investigation of this event is warranted.


7 News Melbourne [7NewsMelbourne] 2016, ‘Melbourne lashed by wild weather. Wyndham Vale appears to be worst hit suburb. @laurelirving7 has an update.’, Tweet 18 March, viewed 7 April 2016,
ABC 2016, ‘Wild weather causes building damage, fallen trees in 'mini-tornado'’, Australian Broadcasting Corporation, 18 March, viewed 7 April 2016,
Anderson, S [skateanderson] 2016A, ‘A housing estate at #wyndhamvale has been particularly hard hit by the #Melbourne storms @abcnewsMelb’, Tweet 18 March, viewed 7 April 2016,
Anderson, S [skateanderson] 2016B, ‘Residents say it was like a tornado ripped through the streets @abcnewsMelb’, Tweet 18 March, viewed 7 April 2016,
Anderson, S [skateanderson] 2016C, ‘An air conditioning unit was ripped off the roof here. The owners can't find it. @abcnewsMelb’, Tweet 18 March, viewed 7 April 2016,
Anderson, S [skateanderson] 2016D, ‘Some pretty serious roof damage at Manor Lakes estate in #wyndhamvale @abcnewsMelb’, Tweet 18 March, viewed 7 April 2016,
Anderson, S [skateanderson] 2016E, ‘Wow. Fences torn up at #wyndhamvale @abcnewsMelb’, Tweet 18 March, viewed 7 April 2016,
Anderson, S [skateanderson] 2016F, ‘High risk of tripping over roof tiles on the streets here @abcnewsMelb’, Tweet 18 March, viewed 7 April 2016,
Baker, H and Foulser-Piggott, R 2015, Use of crowd-sourced photographs in post-disaster damage assessment, paper presented at SECED 2015 Conference: Earthquake Risk and Engineering towards a Resilient World, 9-10 July, viewed 7 April 2016,,%20FOULSER-PIGGOTT.pdf
Davies, B and Duncan, J 2016, Wild weather across Melbourne and regional Victoria causes havoc, 18 March, viewed 7 April 2016,
Graettinger, AJ, Ramseyer, CCE, Freyne, S, Prevatt, DO, Myers, L, Dao, T, Floyd, RW, Holliday, L, Agdas, D, Haan, FL, Richardson, J, Gupta, R, Emerson, RN and Alfano, C 2014, Tornado Damage Assessment in the aftermath of the May 20th 2013 Moore Oklahoma Tornado, viewed 7 April 2016,
Irving, L [laurelirving7] 2016A, ‘Andrew has lost his trampoline and aircon after freak wind went through Wyndham Vale. #7NewsMelb’, Tweet 18 March, viewed 7 April 2016,
Irving, L [laurelirving7] 2016B, ‘Residents say they've never seen anything like it. They call it a mini tornado. #wyndhamvale #7NewsMelb’, Tweet 18 March, viewed 7 April 2016,
Irving, L [laurelirving7] 2016C, ‘Weather bureau says there was no mini tornado, but there were wind gusts 80-90km hr when front went thru. Widespread damage.’, Tweet 18 March, viewed 7 April 2016,
Karl Kim, Pradip Pant, Eric Yamashita 2015, Disasters, drones, and crowd-sourced damage assessment, paper presented at CUPUM, 7-10 July, viewed 7 April 2016,
Kryvasheyeu, Y, Chen, H, Obradovich, N, Moro, E, Van Hentenryck, P, Fowler, J and Cebrian, M 2015, Nowcasting Disaster Damage, viewed 7 April 2016,
Saban, A 2016, ‘Mini tornado hits Wyndham Vale street’, Star Weekly, 18 March, viewed 7 April 2016,


Post a Comment