Wood bowstring trusses seem to fall into a gap in construction and engineering history. They are definitely not vernacular, in that some level of engineering design was necessary for their construction, but they were also not highly designed, and many details seem to have been left to the carpenters. The picture above is taken from a 1915 textbook called Modern Practical Carpentry. So, having discussed the pros and cons of these roofs, I want to discuss some of the gray areas.
First, the early history of these trusses goes against the grain of US construction history. There is a large collection of literature on the development of stick-built construction in the US as a response to the national conditions in the 1800s of the expense of skilled labor, the abundance of lumber, and the need to ship lumber long distances to construction sites. The development of all-stick construction forms in the mid-1800s was supposed to be driven by US-specific issues, but the fact that the earliest versions of these roofs were built in Ireland, and hence named Belfast trusses, throws a wrench into the works. This form was not much used in the US before 1900, while the Irish examples date back to the 1860s. While current literature usually uses the name “bowstring truss,” it’s quite easy to find the name Belfast in older pieces, such as a 1921 American Lumberman article or a 1911 how-to article in The National Builder. It appears that, at some point between when people in the US started using this form and stopped, we forgot that we had imported it.
Second, many, probably most, of the US examples do not meet the specific requirements for the name “Belfast Truss”, which includes a specific geometry to the lattice web. Most of the bowstring trusses I’ve seen have lattice members running at 45 degrees to horizontal, which is easier to lay out and build than the Belfast geometry. So, for truss terminology purists – a group I tend to agree with more than not – we didn’t build Belfast trusses in the US, we bastardized them.
Third, these are not normal lattice trusses, although they share issues. The majority of lattice trusses built in the US had parallel chords, and so worked as trusses (regardless of whether the web members were actually designed or just represent the carpenters’ guess). Bowstring trusses, particularly those with lattice webs, are in many ways more similar to tied arches than to trusses, specifically including the problem of end thrust if the bottom chord is damaged. A parallel-chord lattice truss is effectively a beam; a bowstring truss is more like a tied arch even if it has warren verticals and diagonals in the web rather than lattice.
Finally, bowstring trusses in the US have a history of catastrophic failure worse than other wood truss roofs. This may be in part because of the way they’ve been misused and badly altered over the years, but Dan Eschenasy makes a good case that they were simply misdesigned from the beginning. The kind of mistake he describes, where the chords were sized using the wrong allowable stresses, is the kind of thing that usually doesn’t show up immediately. The result of that misdesign is that the safety factor was lower than it should have been, maybe 1.2 instead of 1.5 under full load. Full load rarely happens, so a bad design can hang around for a long time. But if the load is increased (by adding a mezzanine) or the roof is weakened further (by weathering, or by removal of lattice members) the safety factor creeps down, and will eventually reach the point where it’s less than 1.0 under ordinary loading. Or, as appears to have happened in the Waldbaums fire, the construction of a mezzanine overloaded the roof so that relatively minor fire damage caused a sudden collapse.
Reputation is a funny thing. There’s nothing inherently wrong with bowstring trusses with lattice webs (whether Belfast geometry or not) but I have a hard time seeing them ever used again in the US. We don’t trust the bowstring-truss roofs we have, so how would we build more?