When a wall is subjected to out-of-plane forces, intersecting walls can act as buttresses providing the wall from toppling over. This buttressing action accounts for the scooping shape, where the line of failure high close to the buttress, and low farther away. A particularly important condition for the Macellum is one where a wall has no intersecting wall at one end, as shown in figure 5.1; this condition is termed a wing-wall condition.
As with other out-of-plane failure conditions, the example in figure 5.1 shows that the line of failure is higher near the buttress (on the right of the photo). Nearly all of the shop walls in the Macellum are wing walls. This condition makes possible a staged progression of failure where the main wall is initially well buttressed by the shop walls, but the shop walls fail out-of-plane as wing walls, reducing their effectiveness in buttressing the main wall, leading to its subsequent collapse.
The height, length, and thickness of a wall naturally influence its vulnerability to damage. In discussing vulnerability, the length is the horizontal distance between effective buttresses, rather than the overall length of the wall. This distance corresponds to the structural span of the wall for horizontal loading.
The determination of structural height involves more uncertainty than span. As discussed above, the height of the Macellum walls in ancient times is uncertain, as is the nature of the roofs. In addition, the exterior ground level varies around the building. For the purposes of comparison among areas of the Macellum, the study assumes that height conditions are effectively equal.
The determination of thickness also involves uncertainty. There are limited locations where the thickness can be measured1. The damage areas on the north wall were measured at small openings, and the thicknesses of the south wall were measured by a detailed survey of the southeast corner (see appendix D). These measurements found wall thicknesses in the range of 0.50 to 0.54 meters, neglecting plaster coatings.
There are a variety of masonry construction types in the Macellum, varying both in the materials used and their configuration. There is an important question concerning the relative strengths of these constructions; e.g. is an opus reticulatum wall stronger, weaker, or effectively the same strength as an opus incertum wall with all other factors equal? Vitruvius wrote that opus incertum was stronger and less prone to cracking [Vitruvius 1960, VIII-1, p. 51], but this assessment does not provide a firm basis for comparison seismic performance. Lacking such a basis, the study does not consider these differences in drawing its conclusions.
Although the ground motion of an earthquake is often highly random, it can also have a predominant direction of motion. If there is such a predominant direction, then walls oriented perpendicular to that direction will experience higher out-of-plane loads than those oriented parallel. More so than seismic motion, pyroclastic flow can have a strong directionality that exerts much larger forces on walls perpendicular to the flow than on those parallel. Heilprin observed this phenomenon at St. Pierre in 1902 [Heilprin 1903, pp. 27-29].
Since the out-of-plane damage patterns at the Macellum are discernable only in east-west oriented walls, there is a possibility that orientation is a factor. Mt. Vesuvius is located to the north2, so that east-west walls would be more vulnerable to pyroclastic flow; however, as discussed in detail in chapter 6, the lack of visible repair in the north-south walls is more likely the result of other factors. On the east end of the building, the walls probably experience little damage because of their configuration. On the west end of the building, the walls may have been so severely damaged that they were nearly completely rebuilt.
With the exception of Area E (at the southeast corner of the building), all of the damage areas occur in areas where the main wall of the building has shop stalls on the north side. There are several aspects of the framing which are structurally significant, as discussed below.
The pockets in the wall ends are significantly lower than those in the wall faces: the bottom of the end pockets are 2.21 meters above the ground, while the bottom of the face pockets are typically 2.85 meters. The end pockets extend 60 to 80 cm into the wall, suggesting that timber beams were deeply embedded and cantilevered outward, raising the question of what the beams supported, and why they were 0.65 meters lower than the beams supporting the upper floor. Figure 5.3 addresses this question with a proposed reconstruction of a typical south shop, where it is assumed that the beam pockets in the wall ends supported cantilever beams, dimensioned 17 cm wide and 19 cm deep (corresponding to measured dimensions of beam pockets), and that these beams supported perpendicular beams (assumed the same depth as the cantilever beams) topped with planking and concrete topping, forming an upper gallery along the front of the shops3. The reconstruction includes a thick concrete floor topping, similar to that described by Ulrich [1996, p. 139] for the upper floors of Pompeiian shops.
|Figure 5.3: Cross section showing a reconstruction of a typical south shop, including a gallery and second floor framing.|
If the combined depth of the planking and concrete topping is the same for both the gallery and the second floor, then the difference in height between the two is approximately 53 cm. The pilaster at the end of the shop wall dictates a horizontal separation of approximately 58 cm. Therefore, the two levels can be connected by a stair with three risers of 17.7 cm and two treads of 29 cm: very reasonable dimensions for stairs. This reconstruction accounts for the difference in height between the beam pockets for the gallery and the floor framing.
In addition to accounting for the locations of the beam pockets, the reconstruction also explains why the south shop floor framing does not include an opening for stairs. As shown in the photo of figure 5.2, the beam pockets extend the full depth of the shop from the back wall to the interior face of the pilaster. The connection of the floor to the back wall is clear in Figure 5.4, which shows the gap in the plaster where a beam was against the shop back wall.
|Figure 5.4: Overview of the interior face of the south wall. The line of the second floor framing (highlighted in the lower photo) is clearly marked in the remaining plaster.|
Figure 5.5 shows a perspective view of a reconstruction of the floor framing and gallery. Note that the leftmost shop in the figure, shop 50, does not include floor framing; this is because the wall W50.51 clearly lacks beam pockets. It is possible that this shop housed the stairs used to access the gallery.
|Figure 5.5: Perspective view of reconstruction including south floor framing and gallery. Note that framing is absent from the leftmost shop, the leftmost shop wall lacks beam pockets.|
This configuration is quite different from the north shops, where evidence shows that the second floor framing included an opening in the back of the shop to accommodate stairs. Figure 5.6 shows W16.17 and W17.16 separating shops 16 and 17 near the northeast corner of the building. Close inspection of the beam pockets reveals that there is not a pocket adjacent to the main wall, as there is in the south shops, thus leaving a gap of roughly 50 cm between the back wall and the last beam pocket.
|Figure 5.6: Beam pockets in north shop wall. Upper: W17.16; lower: W16.17. Note that there is not a beam pocket adjacent to the main wall|
Evidence in shop 4, also on the north side of the building, shows that this gap provided an opening in the floor to accommodate a stair from below. Figure 5.7 shows the interior of shop 4, where the back wall includes a prominent diagonal strip of plaster, terminating in the vestiges of a stair landing at its base. This diagonal strip corresponds to the slope of a timber-framed stairway at the back of this shop.
|Figure 5.7: Evidence of stair framing in shop 4 (the photo has been manipulated to show more detail in dark areas.)|
The stairs at the back of the north shops are significant because the opening in the second floor interrupts the connection between that floor and the main wall. Figure 5.8 shows a proposed reconstruction of a typical north shop, with the likely configuration of the stairs and floor framing.
|Figure 5.8: Cross section showing reconstruction of stairs and second floor framing at typical north shop.|
There is a question whether the stair included a landing at the top, which would provide a partial connection between the floor and wall, however there is evidence that stairs in Pompeiian dwellings did not necessarily include such landings. Figure 5.9 shows a stair in a house near the Macellum (VII-IV-36). The stair terminates at its top without a landing, abutting into the wall. The beam pockets adjacent the top of the stair indicate the position of the floor framing.
The northern shops of the Macellum may have used a similar configuration, with a landing at the bottom of the stair but not at the top, so that the stair opening prevented any connection between the floor framing and the main wall. Comparing the seismic response of the typical north and south shops reveals the significance of the stair openings.
The key concept of seismic response is that the earthquake accelerates the ground, which in turn accelerates the mass of the building, inducing forces of inertia. The building mass is located not only in the thick masonry walls of the Macellum, but also in the second floor framing, which were likely topped with 12 to 30 cm of concrete [Ulrich 1996, p. 139]. Figures 5.10 through 5.12 show three-dimensional diagrams which illustrate the effect of ground motions in different directions.
|Figure 5.11: When the ground moves to the north, the building lurches to the south. The inertia of the floor is resisted by the shop walls, oriented parallel to the ground motion.|
When the earthquake accelerates the ground to the north, the building lurches to the south. The inertia of the wall is resisted primarily by the shop walls because of their north-south orientation. The inertia force is transferred from the floor to the shop walls primarily through the floor beams that penetrate the shop walls, and probably also through the contact of the concrete topping with the shop walls. Because it is oriented perpendicular to the forces, the perimeter wall provides little resistance to the inertia load of the floor.
If the ground accelerates to the west, then the roles of the walls are reversed, as shown in figure 5.12 below, with the perimeter wall providing the primary resistance. Note that in an actual earthquake, the motions are typically highly variable, although sometimes have a predominant direction of motion. The key for the south shops is that the structure has a clear mechanism of load resistance for forces from any direction.
|Figure 5.12: When the ground moves to the west, the building lurches to the east. The inertia of the floor is resisted by the main wall, oriented parallel to the ground motion.|
In the case of motion parallel to the perimeter wall, the transfer of the floor's inertia is solely through the contact of the concrete topping, making it much less secure than the connection to the shop walls, which also includes force transfer through the penetrating floor beams. The force transfer at this connection would result from interlocking of the topping and wall materials created when the concrete topping hardened against the rough surface of the wall. There is a question whether such a connection would be able to sustain a major earthquake. Evidence in Area D, discussed in the chapter 6, suggests that much of the south shop construction predates the 62 earthquake, indicating that the connection was adequate.
In considering the north shops, the response to north-south motion is essentially the same as that for the south shops, with the shop walls providing the primary resistance, but the situation is very different for east-west motions. The stair opening at the back of the shop severs the connection between the floor and the main wall, so that the floor's inertia must be resisted entirely by the shop walls, which are oriented perpendicular to the load, leading to a series of failures that eventually collapse most of the shop walls and the main wall. Figure 5.13 shows a diagram of typical north shops, and figure 5.14 illustrates the effect of a westward ground motion.
|Figure 5.13: In the north shops, the floor framing does not connect to the main wall because of the stair opening.|
|Figure 5.14: The inertia of the floor can be resisted only by the shop walls, which are very weak in resisting out-of-plane loads. The floor can break through the walls like a battering ram.|
With the inertia of its massive concrete topping, the floor effectively breaks through the shop walls, like a battering ram. The pattern of failure illustrated in figure 5.14 is a rough estimate, rather than the result of structural analysis, but is consistent with observed patterns of failure in masonry walls
|Figure 5.15: Little remains of the shop walls after the collapse of the upper floor, so that the main wall loses the buttressing of the shop walls|
Note that the failure of the shop wall could leave large portions of the main wall standing unbuttressed, making the main wall vulnerable to north-south motions, as shown in figures 5.16 and 5.17 below.
|Figure 5.17: At the end of the process of progressive collapse, both the shop walls and main walls have suffered serious damage.|
The lack of stair openings in the south shops clearly contributed to their better seismic performance. In the north shop, the effects of the stair openings appear to have been significant in some areas, but mitigated in others by other structural features, discussed in detail in chapter 6.
The span of the floor beams and the nature of their connection to the masonry walls influences their capacity to carry gravity loads, which is an important consideration for events in 79, since the first eighteen hours of the eruption included a rain of pumice stones on the city, totaling a depth of approximately 2.5 meters and imposing significant gravity loads on the timber framing. It is possible that the weight of the pumice could have collapsed timber roofs and floors, making the walls more vulnerable to damage resulting from ground motion.
Sigurdsson [1985, p. 351] concluded that roofs began to collapse with an accumulation of approximately 40 cm of pumice, however this probably underestimates the strength of typical Roman construction. There were two types of pumice fall: an initial layer of 1.3 to 1.4 meters of white pumice, followed by 1.1 to 1.3 meters of denser gray pumice. The average diameter of the pumice fallout was 1 cm. According to measurements by Carey [1987, p. 309], 40 cm of the initial pumice fall corresponds to a load of approximately 250 kg/m2 (51 lb/ft2), and the weight of the full 2.5 meters of pumice would be approximately 2330 kg/m2 (476 lb/ft2).
Although there is little evidence to determine whether the Macellum roofs could have sustained such loads, it is clear that the south shop floors could have sustained it. Based on the spacing and dimensions of the beam pockets, these floors could have carried loads exceeding 3000 kg/m2 (615 lb/ft2) at stress levels considered safe for modern timber; stresses required to induce failure would be more than double.
In the north shops, it is much more difficult to assess the strength of the floors because so few of the shops retain beam pockets, and many of the pockets that do remain appear to be modern reconstructions. One clear factor is that some of the north shops have much longer spans, such as shops 7 and 9 which have spans exceeding 4 meters. These two shops are also particularly vulnerable because they are adjacent the gate, which means that the beams terminate at the shop rather than continuing across to and adjacent shop. This condition reduces the beams load-carrying capacity.
Failure of the north shop floor framing due to pumice fall would certainly change the seismic behavior of the system, although the effects are not completely clear. The collapse of the framing in some shops would reduce limit the scope of damage resulting from the progressive collapse mechanism described above, however the collapse of the roof and floor framing would remove any bracing effect provided by the damage, leaving the shop wall vulnerable to out-of-plane failure, which would then leave the perimeter wall unbuttressed and vulnerable. It appears that pumice induced collapse may have influenced the damage in shop 7, this is discussed in detail in chapter 7.
The preceding discussion identified the following five factors contributing to the vulnerability of walls to seismic and volcanic damage:
Of these, intersecting walls and horizontal framing play the most significant roles in this study. Concerning dimensions, the thickness of the main north and south walls are nearly consistent around the building, at approximately 50 cm, and the height is assumed consistent at 6 meters. The only significant dimension for comparing walls is the length, defined as the horizontal distance between buttressing walls.
Concerning horizontal framing, stair openings play a key role in the behavior of the structural system, since a series of openings in adjacent shops can sever the connection between the upper floor and the main wall, leading to a progressive collapse where the inertia of the floor first collapses the shop walls, leaving the main wall unbutressed and vulnerable to collapse. The gravity-load capacity of the horizontal framing may also have been significant because of loads created by pumice fall in 79 AD. The floors of the south shops were probably strong enough to carry these loads, but some of the north shop floors were weaker because of longer spans.