Friday, April 3, 2015

Moore damage survey of March 25, 2015

The Moore tornado of March 2015 provided a challenge to damage surveyors to derive an accurate rating.  Initially the NWS Norman office gave a preliminary rating of EF1 which led to some heartache by some of the media and yet allowed the survey teams to reassess the tornado strength.  Eventually the members of the four survey teams met to discuss whether or not the tornado should be upgraded.  There were several areas that were under consideration for an upgrade and the teams decided that the damage to several structures was enough to assign the tornado an EF2 rating and others that did not make the cut. This post provides some of the thoughts that went into deciding what to rate the tornado based on these structures.  Before going on, you may wish to visit the NWS Moore tornado page for detail on the tornado path and also this post about the meteorological aspects of the tornado.  I also highly recommend this blog entry by Robin Tanamachi for a well written story of what what it was like to be surveying this tornado.  This blog post is a more detailed perspective from someone who is a meteorologist trying to understand the point of view of an engineer and a forest ecologist when faced with damage to structures and trees.  I try my best but both the engineer and the ecologist have a much more in-depth understanding of the terms I attempt to use and the application to their fields.  You may find some edits to this post once they begin to read it.  :^)

The Moore tornado began at 2335 UTC on March 25 near Southwest 119th St. and May Ave. in southwest Oklahoma City and ending near Northeast 36th Ave and Indian Hills Road.   As the tornado matured it moved east and then took a turn to the southeast.  The track was surveyed by four teams:  The first covered the beginning quarter of the track (Chris Spannagle, Steve Mullens, Alyssa Bates and Matt Taraldsen), The second looking over the wide portion of the track just west of I-35 included Doug Speheger, Tiffany Meyer and Robin Tanamachi, The third covered the area from I-35 and east for three miles involved Dan Dawson, Robert Prentice and I, and then a team of one, Greg Stumpf, surveyed the final few miles of the track.
Figure 1.  A map of the Moore tornado of March 2015 including damage tracks and also a faded image of the rotation tracks where red indicates azimuthal vorticity of .01 /s or greater.   The three boxes represent areas of 

The first area of interest included the only areas of damage determined to be EF2.  These were in the vicinity of the Southgate elementary school area and then a half mile to the southeast.  Both team one and team 2 surveyed these areas but Robert Prentice and I did a more detailed survey of the Southgate elementary school (labeled '2' below).  The area of interest to me was the house with the highest degree of damage found in the entire tornado track and that is labeled '1' in figure 2.
Figure 2.  An intermediate zoomed view of the most intense portion of the tornado with EF scale contours and individual damage points where blue is EF0, green is EF1 and yellow, EF2.

The Southgate school area contained the most intense portion of the tornado with the maximum width and intensity.  Much of the tornado was dominated by a northwesterly inflow jet from the RFD with occasional corner flow vortices.  The strongest of these vortices hit just west of Southgate elementary and then to the southeast.  I’ll focus on the first one that included Southgate and a couple houses to the west (labeled 1 and 2 in figure 3).
Figure 3.  The highest zoomed view of the first area of interest including the EF2 house (1) and the Southgate elementary school (2a-c).  
 The house in the images below lost several walls once the roof was lifted away (figure 4).  According to the teams surveying this house, there were anchor bolts driven every 8’ to a depth of about 2” along the bottom sill plates into the foundation.  Otherwise there were shot nails driven into the foundation.  Not much is known about the roof-to-wall connections, however the house appears in Google’s street view product and some assessment can be made about the shape of the structure (figure 6).  While the house was embedded in shrubs a gable roof can be identified both on the west and south ends.   Since the debris in the picture facing NW appears to have blown eastward, the west gable was likely exposed to the strongest winds.  It is possible the weak links in the house were the gabled west side of the roof and the garage door seen in the second picture (figure 5).  The expected wind speed for collapsed exterior walls was 132 mph but team 2 lowered the wind speed down to the lower bound 113 mph.  Both lie within the EF2 rating.  The Cleveland County Assessors site indicates that the house was built in 1963. 

The next house to the north also lost its roof though it preserved all of its exterior walls.  The surveyors lowered the estimated wind speed to 107 mph, which is EF1, based on their similar appraisal of weaker than typical construction quality.
Not many other DIs exist in close proximity of these two houses, except for other upstream and downstream houses.  The trees in the pictures were located outside the core region of the tornado thus leaving only the telephone poles as potentially representative confirming damage indicators.  All of them were replaced after the tornado, however a piece of one lay against one of the collapsed walls in the first picture (figure 4).  Broken wooden poles also yields an EF2 rating with an expected wind of 118 mph.

Figure 4.  A photograph of the EF2 house with roof gone and some walls collapsed.   
Figure 5.  A photograph of the same house from its west side facing east.  The garage door blown inward was below a gabled roof end.  The house on the left also lost its entire roof but all walls remained standing.
Figure 6.  A Google street view picture of the house from the same view as the first image of the house above.  

 The Southgate elementary school, built in 1963, lost part of its roof on the southwest end of the building.  This image in figure 6 shows some of the debris that landed to the west of the school consisting of insulation boards and roof membrane.  

Figure 6.  The Southgate elementary school viewed from the west (labeled 2a in figure 3).
The southwest side of the school featured a peculiar roof uplift forming an arch, upon which the roof soffit was tilted into the vertical and may have subsequently acted as a vertical load support (figure 7).  Whether this support prevented the collapse of the roof deck is possible though unknown at this point.

Figure 7.  The school viewed from the southwest (labeled 2b in figure 3).
 The interior view of the roof uplift revealed in figure 8 shows an unreinforced CMU block wall.  In this hall the weak link was the grouting one course below what appears to be a very thin bond beam.  In the next room to the east the bond beam separated from the steel web joists.  The interior also shows the soffit that tilted vertically and helped to possibly keep the roof supported through its studs and plywood sheathing.  It’s not known here whether the uplift pressure was from  Bernoulli forces above the roof edge or interior pressure from breached windows in adjacent classrooms.  More likely a combination of the two forces pushed the roof upward.  The roof covering was still intact in this picture and shows insulation boards attached to the web joists.  I was surprised to see no metal roof decking.  Thus I considered this roof to be lower than expected quality for a wind speed estimate.

Figure 8.  An interior view of the uplifted roof over the entrance visible in figure 7.
Two rooms to the northwest a classroom lost all the roof decking, most likely contributing to the debris field to the west of the school.  No metal roof decking was visible in the debris field (figure 9).  Though here the bond beam appears to be a metal girder, stronger than the masonry bond beam further southeast, I considered this roof to be lower than expected quality here as well due to the lack of significant roof decking.  I rated this portion of the school as uplift of roof decking and significant loss of roof covering (>20%) with an expected wind speed of 100 mph.  I consequently lowered the wind speed to 95 mph.  This estimate still yields an EF1 rating.

Figure 9.   An interior view of a class room on the west side of the school.
A second area just east of I-35 had a single story house that lost the entire roof (figure 10).  The team in which I was on inspected this house to determine which side of the EF1/EF2 boundary the house should be rated.  Every house in the expected damage track was rated but only those appearing with colored triangular icons exhibited damage.  Thus the vortex was mostly too weak to cause damage except to a few houses where a brief damaging subvortex may have occurred. 

Figure 10.  An overhead map of the area of interest near Broadway Ave. just east of I-35.  The inset left points to the same area as the white outlined box in the main map.  The circle in the inset refers to the house in question.

This image in figure 11 shows the roof debris blown to the south side of the house consisting of composite shingles, 1by4” decking boards (likely tongue and groove) and 2X4” rafters spaced 24” apart.  Note the adjacent house to the east exhibited no apparent damage.   Adjacent houses to the west and north also showed no visible damage.

Figure 11.  A view of house labeled photo 2 in figure 10 facing northwest.   Figures 12 and 13 were taken at the label '1a and 1b' in the photograph. 
 This image on the east side of the house shows the part of the gable wall top plate and 2 nails (6d?) separated every 18” approximately (figure 12). 

Figure 12. A view of the top end of the collapsed gable roof including the gable wall top plate.  

The rest of the gable showed two nails between each rafter and the gable top plate approximately every 18”  (figure 13).  The gable fell nearly intact, however the orientation of this side was nearly parallel to the strongest wind, as determined by the direction of debris. It appeared though that the gable was poorly connected and exhibited no bracing.  While the lack of bracing appears to be common, the lack of a strong connection also appeared to be less than typical quality construction.

Figure 13.  A view of the rest of the gable.

Perhaps more significant than the low construction quality of the deroofed house was the lack of damage outside that caused by the debris.  These surrounding houses to the southwest showed little to no damage and even the fences remained standing outside the debris zone (figure 14).  The backyard shed was tipped over but it appeared to have no connection to the foundation other than gravity.  The light poles also were undisturbed except where the wires intercepted blown debris.  The only damage in the immediate vicinity occurred from impacts from debris.  As the debris from the roof and a car port blew to the southeast, it struck the fences and another house to the southeast resulting in the loss of exterior brick fa├žade and several windows (figure 15).  
Figure 14.  A view to the southeast from the house.  

Figure 15.  Two panoramas where the top one faces the house from the south end of its backyard.  The bottom panorama was taken from the backyard to the southeast of the house labeled '2' in figure 10.

Why did only this house suffer a loss of roof with the surrounding houses showing negligible damage?  One possibility is that this is the only house to have a car port in the vicinity (figure 16).  So there are two possibilities that affect our decision-making:  1)  The car port debris weakened the roof upon impact causing its failure. 2) A subvortex was small enough to only affect this house to the exclusion of all others.   Both possibilities yield a lower than expected wind speed because either collateral damage could’ve lifted off the roof at relatively low wind speeds or the duration of the subvortex over any spot would've led to a less than 3 seconds.   Consider that the vortex signature on the WSR-88D was moving between 40 and 45 mph and its diameter of maximum winds would be only 50 ft, the maximum winds would not have lasted the standard three seconds defined in the EF-scale.   This vortex would’ve needed to have spun up immediately northwest of the house because there was no visible damage to the houses across the street to the northwest.  However one house to the southeast (not the one with collateral damage) suffered loss of roof deck panels, suggesting the subvortex continued on to the east-southeast.

The degree of damage was labeled as large sections of roof removed, walls remain standing but the three second wind speed was lowered to roughly 105 mph or EF1.  The most likely scenario of what happened here is relatively weak connections between the top wall plate and the roof, along with possible impact from the car port allowed the subvortex to remove the roof from the house.
Figure 16.  A pre- and post-tornado view of the deroofed house as viewed from the northeast.  The pre-tornado view courtesy of Google Streetmaps.
 The final damage indicator that straddled the EF1/2 boundary was in east Moore where a 3600 sq ft house was directly hit by the core of the tornado (figure 17). 

Figure 17.  An overhead view of the final area under consideration, ShadyCreek St. just north of 19th in east Moore.  The inset shows the photo locations in subsequent figures including a wide panorama.
The house is a large single story wood framed structure with a hip roof design, brick and stone veneer and a composition shingle roof on OSB decking.  All the walls were constructed with 2X6” pine studs and roof rafters.  The owner built the house in 2014 and applied updated buildings codes that he promoted to the Moore City council following the May 20, 2013 EF5 tornado.  He mentioned that the house was built with anchor bolts and metal clips from the sill plate up to the top plate. 

The house experienced significant loss (>20%) of roofing above the complex roof-to-wall connections including the decking and numerous rafters (figure 18 and 19).  However there was no damage to windows or the garage doors on the north side of the house.  The adjacent garage shop west of the house lost shingles (DOD=1) and underlay on the south side while the neighboring house lost some shingles (DOD=1) on its west side (figure 18).

Figure 18.  A view of the three buildings including the house with the damaged roof.  Two insets indicate a pre- and post tornado view of the house.  

Figure 19.  A zoomed in view of the west side of the house.

 The tornado passed through the shelterbelt west of the house felling two large multiple trunk Elm trees with trunk diameters approaching 2’ (figure 20).  The smaller trees on either side were not affected thus supporting the premise that larger trees suffer disproportionately from severe winds even in an unfoliated state.   The tree damage indicators in the EF-scale were difficult to apply when vastly different tree response occurs within a just a few meters.  I used the standard wind speed of 100 mph based on either an uprooted or snapped tree trunk and assumed the tornado was of similar strength as at the house.  A narrow section of felled trees also could be seen in the woods opposite the field to the northwest.  The narrow width of felled trees strongly supports the idea of a narrow vortex progressing across the field and over the most affected house.  Due to its small diameter the buildings on either side escaped with relatively little damage.

Figure 20.  A panorama (labeled 'pano' in figure 17) including the houses and the shelter belt to the west.
 With the rafters clipped to the top plate and the lack of damage in the lower roof, why was the upper roof as damaged as it was?  A closer inspection of the roof near the ridge board revealed that the OSB decking was stapled to the rafters at irregular intervals (figure 21).  We could not measure the interval between staples but they were almost certainly spaced at greater than 6” intervals.  Since the building envelope was not breached it’s likely external pressure forcing overwhelmed the limited attachments of the roof decking.  The standard composition shingles offered no more than standard resistance to the wind.  This is below the quality of typical construction to give an expected wind speed and thus the wind speed estimate required to do this damage fell short of an EF2 rating.   Had the decking been clipped to the rafters this roof could’ve survived and EF2 intact and thus similar damage would’ve resulted in a much larger wind speed estimate.  Upon talking to the owner at the time, he admitted that he paid a lot for this flaw in his design.  If he used the same materials and techniques used to construct the porch overhang in framing the roof, he may have saved himself a lot of grief (see figure 22). 

Figure 21.  A zoomed view of the roof ridgeline  labeled '3,4' in figure 17 showing staples used to fasten OSB roof decking to the rafters.  

Figure 22.  A view of the east side of the house.

This survey though is about more than a single rating for the tornado.  With the advent of the Damage Assessment Toolkit (or DAT), we were able to create a high resolution survey from which the whole track could be contoured by EF scale rating.  The map that Doug Speheger of the NWS office in OUN, and others, created from this survey easily shows how small an area the EF2 winds occupied compared to the whole area that the tornado covered (figure 23).   By far the majority of the area covered by this tornado was only EF0 in strength and thus the energy that this tornado exacted upon the landscape was not as big as its EF2 rating suggests it may be.  

This is where high resolution surveys provide their biggest benefit.  Now we can see what fraction of the tornado path is occupied by winds of various strengths instead of making assumptions that has been standard practice based on a small sample of  well surveyed tornadoes.  Now the NWS has the opportunity to vastly increase that sample size and possibility to eliminate the need for more assumptions.  This information is more useful to NWS partners than many of us think.   Considering the users that include insurance industry, the Nuclear Regulatory Commission, risk analysts, FEMA, state and local governments and more and you may begin to see the value that comes from this level of detail in this NWS product.  It doesn't take a team of 12 people for every 12 miles to create them either.  It can be streamlined to be done quickly.  I'll reserve another post for that.  In the meantime, keep them coming!
Figure 23.  The map of the tornado path contoured by EF scale  published by NWS Norman.


To clarify, the media didn't force a reassessment of the tornado rating. Rather the initial rating of the EF1 allowed the NWS time to make a better assessment while still giving the public a preliminary assessment.