Northline Analytics Ep.05 — Where the Building Let Go First: NIST's Joplin Findings on Roof-to-Wall Connections
Every failure sequence NIST described in its Joplin investigation started in the same place: a connection. Not a beam, not a wall panel, not a column. The fastening between one component and the next.
That finding travels a long way from a tornado, because the buildings NIST was describing — light metal roof deck on open-web joists, bearing on concrete tilt-up or masonry walls — are the most common commercial building type in the Treasure Valley.
The derived figure
Using NIST's own numbers, the wind pressure at the tornado's modeled peak was roughly 2.1 to 2.3 times the pressure those buildings were designed for. At NIST's stated upper bound it was roughly 3.1 to 3.3 times. Wind pressure varies with the square of wind speed, so NIST's modeled peak of 175 mph against the 115 to 120 mph code-level design wind speeds it cites gives (175/120)² = 2.13 and (175/115)² = 2.32. At the 210 mph upper bound: 3.06 and 3.33.
Roughly double the design pressure did not break members. It unzipped connections. That is the analytically interesting part, and it is the part that transfers to buildings that will never see a tornado.
The investigation
On May 22, 2011, a tornado moved through Joplin, Missouri. NIST deployed a National Construction Safety Team and published its final report, NIST NCSTAR 3, in March 2014. The report contains 47 findings and 16 recommendations. It remains one of the few published, closed, final federal investigations into how ordinary commercial buildings behave when they are loaded past their design basis.
The tornado killed 161 people. NIST determined that 135 of those deaths, 83.8 percent, were related to building failure. We state that once because it is the reason the investigation exists, and the rest of this piece is about the buildings.
What NIST modeled, and what it did not measure
The storm carried an EF-5 rating on the Enhanced Fujita scale. That rating is derived from observed damage, not from a measurement.
NIST modeled the wind field independently, fitting a Rankine vortex model to tree-fall patterns, and estimated maximum wind speeds of 175 mph with an upper bound of 210 mph — reported as 175 ± 35 mph, with roughly 80 percent of the uncertainty attributed to the tornado's translation velocity. Those two statements reconcile exactly: 175 + 35 = 210.
NIST also found that existing indirect methods carry considerable uncertainty when used to estimate wind speeds for structural design, and that approximately 40 percent of the fatalities and as much as 90 percent of the tornado area were associated with EF-3 or lower wind speeds. Most of the footprint, in other words, was not extreme.
The failure sequence
NIST's investigation team described the collapse mechanism for what it termed box-type system buildings — the umbrella that covers precast tilt-up and concrete masonry unit walls carrying a light steel roof deck. The sequence was consistent:
Wind uplift acts on the roof. The roof deck-to-joist or joist-to-wall connection fails. The roof is what braces the perimeter walls laterally, so losing it removes that bracing. The walls, now unbraced, rotate at the base and come down.
The order matters. The wall did not fail and drop the roof. The roof connection failed and took the wall with it.
NIST identified two features common to the box-type buildings that sustained structural collapse: light-gauge metal roof systems, and friction-only wall-to-footing connections. It noted that friction-only wall-to-footing connections were accepted practice for areas with low or no seismic risk. That parenthetical is important. NIST was describing a design convention, not a defect.
What stayed up
Buildings with redundant lateral load capacity, or that did not depend on roof bracing — steel and concrete moment frames — withstood the tornado without collapse. So did buildings with reinforced concrete or composite concrete-steel roofs. Pre-engineered metal buildings sustained significant envelope damage but no collapses of the primary rigid steel frame.
The difference was not strength. It was whether the structure had a second way to resist lateral load when the first one was removed.
What the design basis actually was
NIST was explicit that conventional buildings are not designed to withstand tornado hazards, and that current codes and standards do not require it. It also stated that available design information showed the roof connections of these buildings were adequate for code-level design wind pressures, making it unlikely they could have failed at wind speeds under the 115 to 120 mph code-level design wind speeds.
So this is not a story about buildings failing below their design basis. They failed above it — but far below the storm's peak. Applying the same velocity-pressure relationship at 136 mph, the lower bound of EF-3, gives 1.28 to 1.40 times design pressure. A 28 to 40 percent overload was enough.
The other derived figures
NIST's investigation team reported 7,411 residential and 553 non-residential buildings damaged, a total of 7,964 — consistent with the "approximately 8,000 buildings" cited elsewhere in the record. From that, non-residential buildings were 6.94 percent of the damaged stock.
NIST selected approximately 25 commercial and institutional buildings for on-site performance surveys. Against 7,964, that is 0.31 percent — a deliberate, small, engineering sample rather than a census. Everything above rests on 25 buildings looked at closely.
The envelope finding, which is the one that pays for itself
The most transferable observation in the whole investigation has nothing to do with collapse. Every engineered building NIST surveyed — those that collapsed and those that did not — sustained significant envelope and interior damage from wind pressure, windborne debris, and water intrusion.
At the regional medical center, the structural system survived. The building was lost anyway, because envelope failure admitted wind and water and destroyed the interior. NIST found the envelope failure, not the structure, was the primary cause of the complete loss of functionality of that facility.
NIST also recorded a direct comparison: the majority of the impact-resistant windows on one floor of the West Tower remained intact, while most of the regular dual-pane insulated windows at the same facility broke under the same conditions. And it found that roof aggregate contributed to envelope damage at that facility and surrounding structures, which is why one of its recommendations is that aggregate surfacing and ballast be prohibited on buildings of any height in tornado-prone regions.
A building can keep standing and still stop working.
What NIST recommended about load path
Recommendation 6 in NCSTAR 3 asks for "the development of risk-balanced, performance-based tornado design methodologies such that all building components and systems meet or exceed the same performance objectives when subjected to tornado hazards."
Stated plainly: stop designing buildings whose parts have mismatched capacities. A wall panel engineered for a load the connection above it cannot deliver is a building with a designed-in weak link.
NIST's Recommendation 12 also asks owners and operators of existing critical facilities in tornado-prone areas to perform tornado vulnerability assessments covering backup power supplies, elevator equipment and shaft enclosures, and means of egress illumination — its backup power finding was that ancillary buildings housing the generators collapsed, taking the redundancy with them.
Why this matters in a valley with no tornadoes
Southern Idaho is not tornado country. The transferable content is not the hazard, it is the load path.
The box-type building NIST described is the standard Treasure Valley industrial, warehouse, and big-box retail form. Wherever the roof diaphragm braces the walls, the roof-to-wall connection is a structural element, not a detail. Uplift comes from ordinary wind too, and roof-mounted equipment, added openings, re-roofing over existing decks, and corrosion at fasteners all act on the same connections NIST identified.
Those connections are frequently observable. At the wall-to-roof interface, ledger conditions, embed plates, joist seats, and deck fastening are often visible from inside a warehouse where there is no ceiling. Corrosion, distortion, missing or backed-out fasteners, and previous modification are all things a field observer can see and photograph. A condition assessment does not calculate capacity — that is engineering — but it can flag the connection that has been quietly modified, or the deck fastening that has corroded, and put an engineer on the right question.
Limits of this analysis
The pressure ratios are ours, derived from NIST's published wind speeds using the standard relationship that wind pressure varies with the square of velocity. They are an order-of-magnitude comparison, not a design calculation: real tornado loading involves atmospheric pressure change, debris impact, directionality, and internal pressurization that no single ratio captures.
The 115 to 120 mph design wind speeds are NIST's general characterization of code-level design at the time, not the design basis of any particular building.
The full text of NCSTAR 3 could not be retrieved in this analysis. The report identity, finding count, and recommendation text above come from NIST's publication record and from NIST's own published extract of the final report's recommendations. The failure-mechanism findings come from NIST investigation team presentations to the National Construction Safety Team Advisory Committee dated December 10, 2013, which present the findings at draft stage; the final report contains 47 findings and its numbering differs. We have therefore described the substance and not cited finding numbers.
The building counts come from those same presentations rather than from the final report, and NIST's own documents give slightly different roundings of the residential total in different places.
The sample is 25 buildings from one event in one city, surveyed by engineers who selected them precisely because they were informative. It is not a random sample of anything.
And this is a tornado. Nothing here should be read as a statement about tornado risk in Idaho.
Frequently asked questions
Does this mean tilt-up buildings are unsafe? No. NIST's finding was about a load path, and it described friction-only wall-to-footing connections as accepted practice in low-seismic regions rather than as a defect. Tilt-up and masonry box-type buildings perform well within their design basis. The finding is about what happens past it, and about which element gives first.
Can an inspection tell me whether my roof-to-wall connections are adequate? No. Adequacy is a structural engineering determination. What an inspection can do is document what is visible, identify corrosion, modification, or damage at those connections, and tell you when an engineer is warranted.
Why does the envelope matter as much as the structure? Because functionality is what a building is for. NIST's clearest single lesson from Joplin is that a facility can survive structurally and still be a total loss through the envelope.
