Moynihan Train Hall

January 2022

MIT

Bird's-eye poster of Pennsylvania Station, New York City, with the Farley Post Office in front of it and the city stretching back to the river

At the turn of the 20th century, two large public buildings were built above the train lines in New York: Penn Station and the James A. Farley Post Office.

Photographs of the James A. Farley Post Office Building in 1914 and Pennsylvania Station in 1910

These two Beaux-Arts buildings sit next to each other on Eighth Avenue in Midtown West. Both were designed by McKim, Mead & White.

What started as an open-air station was capped by Penn Station and the Farley building. In 1963 Penn Station was demolished by the Pennsylvania Railroad Company to make way for Madison Square Garden.

Animation of the blocks between Seventh and Ninth Avenues: open-air platforms in 1906, Penn Station and the post office in 1912, the post office annex in 1960, and Madison Square Garden rising on the cleared station site in 1967

Photographs of Penn Station's steel frame during demolition, and of Madison Square Garden under construction

The demolition was a very controversial decision, and it became a turning point for how the US thought about architectural preservation.

Enter Senator Daniel Moynihan

Moynihan held a position at the Harvard–MIT Joint Center for Urban Studies, and in the early 1990s he spearheaded a plan to adapt the Farley Post Office building into an extension of Penn Station, because he was nostalgic for the old Penn Station and his time commuting through it.

Sectional perspective of the train hall, from the tracks and platforms up to the four grid shell vaults, with red marks where the vaults meet the existing structure

SOM won the project with a design in which the old mail sorting room is now a new passenger sorting room, under four large cable-braced steel grid shells that use the existing steel trusses.

Renderings of three losing proposals for the hall, by James Carpenter Design, SOM and DSR

Losing proposals from a range of firms used a similar strategy, some keeping the existing trusses and some replacing them. This led to the question: what are the structural, spatial and embodied carbon consequences of keeping the historic trusses?

For the study, a workflow of form-finding (in Kangaroo), structural analysis (in Karamba 3D) and embodied carbon analysis (in Cardinal LCA) was used. The existing shells were approximated from a photogrammetry mesh of the building: form-finding with loads scaled inversely to their distance from the centroid, Z = α / (distance from centroid), gave a shell that matched the roof in section.

Load scaling for form-finding, as load vectors radiating from the centroid and the resulting shell, beside a section through the photogrammetry mesh with the form-found shell over the existing truss

Two cases were devised to compare against the existing design, the Base Case (Case 1). Case 2 adjusts the form of the grid shell and still uses the trusses. Case 3 follows one of the losing proposals, which did not reuse the existing trusses.

In each case the cross section of the members (8 to 10 cm wide and more than 20 cm deep) and the form could vary, while the surface area was held constant. The structure was judged by its efficiency, Σ|F|L, the axial force in each member times its length, summed, with the maximum axial compressive force as an upper limit. Embodied carbon was counted in tonnes of CO2 equivalent, from the Inventory of Carbon and Energy (ICE) 2019 database.

Results for the three cases, the existing shell on its trusses, the shell form-found with normalized loads on the same trusses, and a new shell on new columns, with Σ|F|L of 83.1, 66.9 and 357.1 × 10⁴ and maximum forces of −643, −219 and −11,698 kN

Interestingly, Case 2, which uses normalized loads across the grid shell nodes for form-finding, performed more efficiently, with a much lower maximum compressive force.

Plans of the reaction forces around the edge of the shell, for the existing form and the form found with normalized loads

A possible explanation for this is the reduced reaction forces out of plane with the truss at its peak, as can be seen in Case 1.

Planarity of the shell, for the existing form and the form found with normalized loads, green where the panels are flat and red where they are not

The shell form-found with normalized node forces also produced much higher planarity across the shell (seen in green).

Stacked bar chart of embodied carbon in tonnes of CO2 equivalent: the existing trusses 1,526, Case 1 1,736, Case 2 1,731 and Case 3 2,219, split into grid shell, windows and columns

Altogether, Cases 1 and 2 embody a similar amount of carbon equivalent. Case 3, which required additional structure, having removed the existing trusses, unsurprisingly embodies more CO2e.

I feel that it is only fair to include the embodied carbon of the discarded trusses in the calculation for Case 3, which means it embodies about 2 kilotons of CO2e more than Case 2. At the 400 g of CO2 an average car emits per mile, that is 5,000,000 miles of driving, or 200 road trips around the earth.