Archeoseismic study of damage in Roman and Medieval structures in the center of Cologne, Germany
- First Online:
- Cite this article as:
- Hinzen, KG., Schreiber, S., Fleischer, C. et al. J Seismol (2013) 17: 399. doi:10.1007/s10950-012-9327-2
- 1.4k Downloads
The causes of damage observed in archeological records or preserved monuments are often difficult to be determined unequivocally, particularly when the possibility of secondary earthquake damage exists. Such secondary damage has been previously proposed for the Roman Praetorium, the governor’s palace in the center of Cologne. Ongoing excavations since 2007 revealed additional damage. The existing ground that has been uncovered and documented extends the affected area to 175 × 180 m. We present a comprehensive virtual model of the excavation area based on 200 3D laserscans together with a systematic analysis of the damage patterns and an improved model of the terrain during Roman/Medieval times including geotechnical parameters of the subsurface. Five locations with different damage patterns, including a Roman sewer, the octagonal central part of the Praetorium, a section with strongly inclined massive walls, a 13 m deep deformed well, a collapsed hypocaust, and damages in the Medieval mikveh are analyzed in detail. We use site-specific synthetic strong ground motion seismograms to test the possibility of earthquake-induced ground failures as a cause for the observed damage. This subsurface model is also used to test the possibility of hydraulically-induced damages by seepage and erosion of fine-grained material from stray sand. Heavy rainstorms can induce a direct stream of surface water through the fine sand layers to the ground water table. Simulated ground motion for assumed worst-case earthquake scenarios do not provoke slope instability at the level necessary to explain the structural damages.
KeywordsArcheoseismology Earthquake Laserscanning Slope stability Seepage Lower Rhine Embayment Cologne
An unequivocal cause of damage observed in archeological records or preserved monuments is often difficult to determine. Various natural causes may result in similar damage patterns (Galadini et al. 2006), and anthropogenic action can also produce similar-looking damage or deformations (Nikonov 1988). If targets of archeoseismological studies are being excavated in an ongoing project, it follows that major anthropogenic actions have not occurred since coverage of the targets. Unless located directly on a fault zone (Ellenblum et al. 1998; Meghraoui et al. 2003; Marco 2008), earthquake ground motions affecting a specific location after the burial of the objects are not generally responsible for significant altering of such buried targets. Preserved monuments, however, which have been visible and accessible at the surface since their construction could have suffered subsequent damage or deformation. In addition, anthropogenic alterations could have occurred throughout the history of the object including restorations, construction of more recent buildings, and architectural modifications to permit a different use (Galadini et al. 2006). Prominent examples of such monuments are the Parthenon in Athens and the Colosseum in Rome. Compared to such large architectonical structures for practical use, grave houses, large sarcophaguses, and other structures of necropolises are less likely to be used for purposes other than their original intent. However, numerous examples show that such structures were often the target of looters, sometimes immediately subsequent to construction, sometimes late in the objects’ history (Hinzen et al. 2010). In cases where both natural and man-made causes of damage or deformation are possible (Guidoboni 2002; Ambraseys 2005; Galadini et al. 2006), both kinds of sources should be parameterized and modeled, thereby necessitating construction of an appropriate parameter space to test the plausibility of competing hypotheses.
The Cologne Praetorium, the palace of the Roman governor of the province of germania inferior, was initially excavated and partially preserved in the 1950s. All modern buildings in this area had been destroyed during WWII aerial bombing and Roman foundations were exposed during the construction of a new city hall. These foundations and remaining parts of the standing walls show severe structural damage, especially concentrated at the central octagonal part of the former building. The remains of the Praetorium include walls from four building phases (I to IV) between the first and the fourth century CE; however, only the best-preserved walls dating from phases III and IV show significant damage.
The early excavators (Precht 1973) assumed slow static settlements due to a weak subsurface and inadequate building technique as the main cause of the breakdown of wall sections. Hinzen and Schütte (2003) argued for a more sudden event that affected the building ground and subsequently the integrity of the structure. They found arguments for secondary earthquake damage, however also pointed out the need to test for possible hydrologic causes.
Since 2007, extensive excavations immediately south of the Praetorium and of a newly excavated section of a Roman sewer north of the Praetorium provided an opportunity to collect data from a hitherto inaccessible section of the subsurface. The current excavation area (Schütte and Gechter 2011) includes the Medieval Jewish quarter of Cologne, which hosted one of the earliest Jewish communities in Europe.
In this study, we (1) present a comprehensive virtual model of the archeological inventory in a 175 × 180 m area that includes the Cologne Praetorium based on detailed laserscan measurements, (2) quantify the damage to selected subsections, (3) use an improved seismotectonic model to calculate site specific synthetic seismograms for deterministic earthquake scenarios, (4) test the static and dynamic stability of the subsurface, (5) simulate seepage during heavy rain storm cycles to test for possible internal suffusion and erosion processes, and (6) review aerial surveillance photos taken after WWII bomb attacks.
2.1 Tectonic and seismicity
2.2 History and archeological setting
The northward extension of the Roman Empire in the last century BCE and the first century CE produced enormous building activity, and some well-preserved buildings from this period exist in central Europe. During the Roman period, building activity boomed. Agrippina, the wife of the Emperor Claudius, made her birth town in 50 CE a colony with the name Colonia Claudia Ara Agrippinensium (CCAA), and the inhabitants became Roman citizens. The name ‘Claudia’ goes back to the family of the ruling emperors, the Claudians; “Ara” refers to Ara Ubiorum, the name of the main altar of the Germanic province, the germania inferior (La Baume 1980). The settlement was located on a plateau of 1 km2 of the Lower Terrace, ca. 14–16 m above the Rhine River and safely above the level of flooding (Gechter and Schütte 2000). To the east, an arm of the Rhine River bordered this settlement. The city walls of the Oppidum Ubiorum were laid out in a rectangular pattern similar to Italian Roman town foundations, while the city walls of the second half of the first century CE followed the irregular pattern of the terrace edges (Doppelfeld 1956). At the riverbank, the city wall (Fig. 1) did not follow the terrace edge but was placed right at the waterline. The investigated structures are located on this terrace and in the area of the slope towards a former side arm of the Rhine River.
A summary of the history of the archeology of the area is given by Schütte and Gechter (2011). Already in the thirteenth century people building a new house opposite the city hall were aware that this new structure was founded on Roman walls. Weinsberg (Höhlbaum 1887) described in 1570 Jewish and Roman foundation walls on the town hall square. Gelenius (1645) published an inscription that gave clear reference to the Cologne Praetorium. During construction works on the town hall between 1861 and 1865, subsurface walls were recognized as those of the Roman Praetorium and Schultze et al. (1885) closely examined these walls, but they interpreted them as Franconian.
Due to the more than 95 % damaged part of downtown Cologne, intense construction work was initiated after WWII. During the excavations for a new administration building at the location of the destroyed “Spanischer Bau,” monumental walls were discovered, and for which the archeologist Doppelfeld (1956) was urged to complete the excavations of in a mere 6 months. Luckily, the City Council decided to preserve the northern part of the new archeological findings of the Roman Praetorium underneath a self-supporting arced reinforced concrete ceiling. Additional excavations mainly south of the Praetorium were made subsequently by Precht (1973) who also gave a detailed description and interpretation of the building sequences. Recently, an open plan for an extension of the existing museum underneath the town hall was revived. In 2007, further excavations started and are ongoing at the time of the writing of this paper. The new excavation area (Fig. 1) includes the Medieval Jewish quarter of Cologne with a synagogue, a mikveh, a hospital and numerous residential buildings.
3 Studied structures
Figure 1c shows a plan of the main structures of the study area, which are from north to south: the northern main sewer, the Praetorium including the octagonal shaped central tower, the so-called Porticus with the Jewish hospital adjoining a massive Roman wall, a Roman well, the Medieval synagogue complex with a large cesspit, a buried collapsed hypocaust, a deformed structure underneath the women’s synagogue; the mikveh with damaged walls in a former entrance, and the residential area with a deformed staircase of Late Medieval. While the western part of the northern sewer and northern two thirds of the Praetorium were already preserved and open to the public, all other structures became accessible during the recent and ongoing excavations for the construction of the Archaeological Zone Cologne. The excavation is unique: it is located right in the center of a megacity and covers 2,000 years of building history. It is in parts more than 10 m deep, includes effects of several pogroms (i.e., 1096, 1349) and wars, especially heavy damage during WWII, and thus constitutes a challenge for archeological and archeoseismological work.
3.1 3D laser scan model
We used the laser scanning technique to create a comprehensive virtual model (CVM) of the remains of the whole excavation complex (Schreiber et al. 2009, 2012; Schreiber and Hinzen 2011). In full resolution, the CVM contains 2.3 billion georeferenced 3D points. Special challenges included the repeated scanning of the 8.3 m deep cesspit during its excavation (Schreiber et al. 2012) and the 13 m deep Roman well (Fleischer et al. 2010). The CVM enhances visualization and quantification of the damage found in the study area, which were previously described (i.e., Doppelfeld 1956; Precht 1973; Hinzen and Schütte 2003), as well as damage described here for the first time in their archeoseismological context. The damage includes (1) cracked foundation walls, (2) subsided structures, (3) tilted structures, (4) opening of horizontal gaps, (5) deformed structures, and (6) a collapsed hypocaust.
The aerial photo in Fig. 1 shows the current building locations in the city center of Cologne together with the outline of the Roman and Medieval structures discussed in the following.
3.2 Northern sewer
The 115 m long western section that is already part of the existing Praetorium museum was completely scanned with 26 single scans. The initial excavation of the eastern section had to be discontinued due to static stability concerns and only an emergency scan campaign has so far been possible. A total of 24 cross-sections through the virtual model of the western sewer section were compared to a master crosscut in order to quantify any deformations. These were mainly found in the tuff block roof (Fig. 2). Several of the northern neighboring blocks of the center stone are dislocated inwards. For each cross-section, the ideal shape of the sewer at this position was reconstructed based on the width of the construction and the height of the sidewalls. Inward and outward deviations of the measured cross-sections from the ideal shape were quantified through the residual areas that measure on average 6.4 ± 2.8 % of the total cross-section. If only the roof is taken into account, the residual area was found to be 9.3 ± 5.3 % of the total cross-section.
The opus caementicium roof of the eastern sewer section rests on walls made from large 0.6 × 1.2 m tuff blocks. A massive crack cuts through the roof, parallel to the trend of the sewer over a length of at least 14 m; it displaces the roof by up to 0.15 m, and the cross-sections from the CVM indicate a clockwise rotation (view to the west) of the northern half of the roof. In addition, the archeologists found several large patches of rock that were detached from the tuff blocks, mainly at the lower part of the walls. These shell-like patches were still in situ and held in position by the sediments, indicating that the detachment occurred after the sedimentation process. Patches of the roof were found on top of the Roman sediments: however, they were not sunken into these layers at all (Schütte, personal communication), also indicating a damaging event after the accumulation of the sediments. The sediments, at least in one section, show signs of reworking after the initial sedimentation (Fig. 2).
3.5 Roman well
The model shows block layer inclination towards east at angles up to 11°, and in addition, the circular well is deformed to an ellipse with the short axis in NW–SE direction. Best-fitting planes for the surface of each building block were calculated and the deviation of the plane normal from the horizontal direction determined (Fleischer et al. 2010). When the 2.5 m sediment fill of the well was excavated in 2008, a wooden construction was found at its base sitting in fine-grained sand. Between the completion of this excavation in 2008 to the time of the lasers canning in 2009, the depth of the well decreased from 13.2 to 12.5 m due to seeping sand.
3.6 Cesspit and hypocaust floor
Following Schütte and Gechter (2011), the antique building in the center of the town hall square (Fig. 6) was in use in early Medieval times prior to the year 800 CE as a synagogue. While the antique building had a central room and an adjoining northern and southern room, the synagogue was reduced to the central room after 800 CE. The walls of the central part were obviously repaired. The southern room was given up; however, the northern room was in use after some repair, indicated by a new floor found in parts of the former northern annex (Schütte and Gechter 2011).
A large cesspit located in the northwestern corner underneath the synagogue was excavated between summer 2008 and spring 2010. It has a diameter of 2.3–2.5 m, and the excavation reached the final depth of the structure at 42.97 m above sea level (asl). The 8.3 m deep cesspit built from 60 to 70 layers of small to medium tuff blocks was statically unstable and the walls had to be stabilized and reconstructed successively when the fill was removed. Schreiber et al. (2012) describe in detail the scanning procedure and implementation of the cesspit in the CVM. At a depth between 49 and 47 m asl, the walls of the cesspit are heavily deformed and damaged. The severity of deformation indicates that only fill in the cesspit prevented the structure from a collapse. The fill is archeologically dated to the middle of the fourteenth century, giving a terminus ante for this damage. At the bottom of the cesspit, the same kind of sand exists which was found behind the cesspit walls and beneath a neighboring profile of a collapsed hypocaust floor.
Large basalt columns were found horizontally layered underneath the arch on the east and underneath the second, lower arch on the west side. These basalt columns were probably not part of the original construction but could have been placed to support the subsided (or subsiding) structures. On top of the damaged section, several floor levels were built at elevations between 47.2 and 47.7 m asl. These date to the time between the eleventh and fourteenth century CE.
3.8 Damage pattern
All damage listed above directly involves or is at least compatible with strong subsurface deformations within the slope and artificial fill in which the structures are found or directly related to fine isolated sand lenses. If they were of seismogenic cause, the damage must therefore be classified as secondary damages as already indicated by Hinzen and Schütte (2003). Typical primary vibration-induced damage (i.e., Galadini et al. 2006; Hinzen 2011) such as corner stone expulsion, shear wall cracking, and toppled walls have not been found in the study area. The damage patterns as recorded in the CVM are either within the range of the slope of the river terrace, i.e., Praetorium, parts of the sewer, the Porticus, the apse with the Roman well, or on top of the terrace, i.e., synagogue buildings with the cesspit, mikveh, and Medieval residential buildings. While the former include mainly tilted constructions, the latter are all primarily consequences of subsidence. However, the block displacements in the western section of the northern sewer, the cracked roof of its eastern section, and the reworked sediments do not fit into the damage patterns listed above.
Source, type, and number of data used for the construction of a digital terrain model of the study area
Heavy anthropogenic alterations of the relief introduce uncertainty into the reconstruction of the original structure. In addition, the auger cores do not always lend themselves to a ready differentiation between naturally deposited and anthropogenic reworked flood loam. The area inside the Roman city wall (Fig. 10) is 47–52 m asl corresponds to 8–14 m above the average Rhine River level at Roman times at 38.86 m (Köhler 1941). This level is only 4 dm higher than the current average level of 38.46 m asl (STEB 2009). The side arm of the Rhine River was connected to the actual riverbed at the northern end. It has been intensively debated in the literature whether a harbor existed in this sidearm of the river at Roman times (i.e., Schultze et al. 1885; Lückger 1919; Köhler 1941; Doppelfeld 1953; Hellenkemper 1980; Wolff 2000). The model presented in this paper does not show a connection to the riverbed at its southern end, and with an average water depth of 1.87–3.65 m, it would have been only accessible by boat from the northern opening. However, accessibility during low-water levels of the Rhine River is questionable if no dredging of the flood loam at the entrance was made.
Material parameters of the seven layers of sediments used in the geotechnical model of the subsurface (Fig. 11)
Specific weight (kN/m³)
Internal friction angle (°)
k (at saturation) (m/s)
Coarse sand 1
Coarse sand 2
5 Earthquake ground motion model
5.1 Seismotectonic model
Seismic moment, moment magnitude, surface rupture length, and strike of 16 earthquake scenarios for the Lower Rhine Embayment (Fig. 12)
5.2 Ground motion
5.3 Slope stability
The 2D FE model (Fig. 11) was used to study (1) the static slope stability, (2) the deformation of the slope during earthquake ground motions, represented by the strong motion synthetics described above, and (3) the possible permanent deformation during shaking including, (4) the time-dependent variation of the factor of safety (FS) for all earthquake scenarios. The calculations were made with the commercial Geostudio 2007 software (GeoStudio 2007). The initial stress conditions were calculated with the ground water table assumed at 38.9 m asl. A total of 729 potential circular slip surfaces were considered, and the range of entry and exit points is shown in Fig. 11. The sliding masses were divided into 30 vertical slices. For each slice, the resulting sum of driving and resisting forces is calculated (Kramer 1996). Slope stability is expressed through the FS which is 1.0 when equilibrium between driving and resisting forces exists and FS > 1.0 if the sliding block is stable. The minimum value for all included slip surfaces is 2.541 indicating stable static conditions far above the limit equilibrium. The slip surface with the minimum FS and a radius of 8.6 m is shown in Fig. 11.
6 Hydrological model
Rainstorm events were simulated by implementing boundary conditions for the model surface based on a rainstorm with a 100-year return period following contemporary models (STEB 2011, personal communication), assuming a similar climate during the late Roman Medieval period. For current conditions, this storm event is estimated with 1.136 × 10−5 m/s of precipitation lasting 3,600 s. As the whole study area was densely covered with buildings and sealed surfaces, we estimate a concentrated inflow of water on the 2D model by taking into account the flow of a 4–10 m wide section.
The Romans installed and maintained an elaborate dewatering system in CCAA. Fresh water was brought to the city by a 100 km long aqueduct from the Eifel Mountains. The constantly running water had to be systematically drained to the Rhine River. The two sewers bordering the study area in the north and the south were essential parts of the dewatering system. Already in the fourth century CE, due to lack of maintenance, accumulated sediments began to block the sewers and after the decline of the Roman Empire, the system ceased to function. Ingenious dewatering canals also existed underneath the Praetorium. It may be assumed that these were also no longer maintained after the Roman period, and the previous well-controlled water runoff of the study area ceased to function.
For a width of this surface larger than 7 m, a significant amount of water reaches even the Quaternary layer within 8 h. The flow through the second measuring section below the loam layer is heavier than through the measuring section above, as a strong flow develops from west to east, down the slope on top of the loam. A part of this flow enters the voids without having crossed the top-measuring surface the same way water flows into a gully. For comparison, the seepage for the three measuring surfaces for a model without the voids is shown for the strongest inflow from a 10 m wide section. Without the voids, basically all water flows down the slope on top of the loam layer, resembling the situation before building construction began and dewatering became uncontrolled. The basic seepage model shows that for a heavy rainstorm, significant amounts of water can penetrate the voids in the loam layer and reach the groundwater level within a few hours.
7 Discussion and conclusions
The Lower Rhine Embayment is a young sedimentary basin with typical intraplate seismicity (Camelbeeck et al. 2007; Hinzen and Reamer 2007). The strongest instrumentally recorded earthquake in 1992 had a MW 5.4. Historical seismicity includes stronger events like the 1756 Düren earthquake (MS 5 ¾ (Camelbeeck et al. 2007)) and the 1692 Verviers earthquake (MS 6 ¼ , (Camelbeeck et al. 2007)). In such a situation where larger earthquakes are rare, any information about former damaging earthquakes is of crucial importance for hazard evaluations. Several palaeoseismic studies in the area have shown that stronger earthquakes than those of the historic record occurred during Holocene times (MW ≥ 6.3, Camelbeeck and Meghraoui 1998; Vanneste and Verbeeck 2001; Vanneste et al. 2001, 2008). The northern Rhine area has been occupied since Neolithic times and building activities began with the arrival of the Romans in the first century CE and lasted four centuries. This situation suggests that earthquakes have left their marks in the form of damage to buildings of the past two millennia. While Hinzen and Weiner (2009) showed that the hypothesis of a seismogenic cause of damages to a Neolithic water well was not plausible even though it was located in close proximity to one of the major active faults of the Lower Rhine Embayment, Hinzen (2005a) found strong arguments for earthquake damage to the Roman city fortifications of Tolbiacum, an important road junction west of Cologne. While time estimates for the damage contain large uncertainties, radiocarbon dating indicates the second half of the fourth century CE.
Possible earthquake damage has also been proposed for a rural Roman building (Hinzen 2005b), for the Carolingian Cathedral of Aachen (Reicherter et al. 2011), and Roman structures in the city center of Cologne (Hinzen and Schütte 2003). The latter include structures studied in detail in this contribution. While Doppelfeld (1956), Precht (1973), and Wolf (2000) argued for static settlements as the cause of this damage, Hinzen and Schütte (2003) argued for a more sudden event. Recent excavations (Schütte and Gechter 2011), new exploratory fieldwork, laboratory measurements, and the CVM of the excavation area offer a comprehensive database to document the cause of manifold damage.
Several structures were damaged later than the proposed time of the collapse of the Praetorium (Fig. 19). The heavily damaged cesspit of the Medieval synagogue was built in the first half of the tenth century CE and the damage likely occurred in the fourteenth century CE (Fig. 19). It is not clear when the walls in the entrance area of the mikveh were damaged; however, 14C dates indicate, that the basin of the mikveh was rebuilt around 1000 CE. Subsided walls of residential buildings in the Jewish quarter and a staircase from Late Medieval times (Fig. 19) are strong indicators for a recurrent instability of the subsurface as opposed to a single, sudden catastrophic event that damaged all structures in the study area.
The date of the damage of the Roman well located within the apse on the town hall square (Fig. 6) cannot be reliably determined. However, the well was repaired in the tenth century CE. Assuming that damage and repair were not separated by centuries, this result is a further argument for several damaging events.
Effects of explosions from WWII bombs are found at several places in the excavation. In the northeastern Praetorium, a direct hit, marked in Fig. 4, destroyed a 5-m long section of a foundation wall. In total, some 150,000 bombs were dropped on Cologne during WWII. The last and largest bombing of Cologne was on March 2, 1945. The aerial surveillance photograph (Fig. 3) taken in April 1945 documents several bomb hits on the eastern section of the northern sewer. Taking into account that the bomb that damaged the Praetorium first penetrated the multistory city hall above it before reaching the underlying Roman foundations, it is feasible that the hit on the road above the sewer caused the structural damage in the sewer, in particular the crack in the concrete ceiling. Assuming a depth of the bomb crater of 2–3 m, the explosion of a direct hit occurred only 4.5–2.5 m above the roof of the sewer. Sewer damage from bombing during WWII is also reported at other locations, i.e., in Newcastle upon Tyne and Sunderland (Ray 1996).
Additional anthropogenic causes are the collateral extraction of fine-graded sand during the water procurement from numerous wells and additional loading to subsurface structures due to younger constructions. The first effect was deemed a feasible cause for the damage to the Roman well and the Late Medieval staircase and its surroundings. The bulging of sand into the Roman well in the apse after it had been excavated is a clear indication of a persisting mobility of wash dirt in the subsurface of the study area. The additional loading from artificial fill and constructions is counted as questionable for the structural damage of the sewer. This results in a plausibility index of p = 0.27. Natural deterioration, or wear and tear, (p = 0.09) is unfeasible for all structures with the exception of the sewer and cesspit. For these two objects, it is counted questionable. Wear and tear could explain the spalling from the sidewalls of the sewer and the cesspit deformation. However, deterioration has not caused the sewer roof deformation. In these cases, wear and tear is rated questionable. While (repeated) flooding from the Rhine River is a feasible explanation for the mobilized sediments in the sewer, flooding is questionable or even infeasible for most of the other damage (Fig. 20) resulting in an index of p = 0.32.
We assigned slow static settlements as a questionable but not completely impossible cause for six of the damage types and feasible for the cesspit. The reworked sediments that filled the sewer and the collapse of the hypocaust under the synagogue were not caused by slow static settlement. Due to the lack of any traces of repair attempts, we consider it also infeasible for the broken Octagon. The index for this damage cause is 0.36.
A seismogenic cause is counted feasible for the structural damage of the sewer as well as the mobilization of its sediment fill and for the collapsed hypocaust. The deformed cesspit is most probably not a result of coseismic damage. For all other structures, the seismogenic cause is rated as questionable because they do not exhibit direct vibration damage but are a consequence of subsurface deformation. In addition, the results of the model calculations showed the stability of the slope and the improbability of liquefaction even during strong ground motions. The p value for a seismogenic cause is 0.59.
With the exception of WWII bombing, the most recent damage (cesspit and Jewish quarter) date to the fourteenth and fifteenth century CE. Covering a time span from the fourth to the end of the tenth century CE, and in case of the sewer up to the twelfth century CE, the damage to the Roman structures (well, apse, sewer, Praetorium, Porticus) are not precisely dated (Fig. 19). The time of damage to the mikveh in the tenth century CE overlaps with most of the long time windows for the damage to Roman structures, however, not with the relative short window for the Praetorium at the end of the eighth century CE. This timeline argues for at least three separate damaging events in addition to the WWII destruction.
While written sources around 800 CE are scarce due to the destruction wrought by the Norman invasions (881/882), it is unlikely that strong damaging earthquakes later than 900 would be missing from the historic records. Ground motions capable of generating the heavy secondary damage in the study area require an earthquake that would have caused widespread damage in and around Cologne and outlying areas. Therefore, even if the occurrence of the earthquake was suppressed in Cologne for political reasons (a natural catastrophe would have likely been interpreted as a heavenly punishment), its effects would have been recorded elsewhere. This absence-of-written-evidence argument also applies to a multiple earthquake hypothesis, at least with respect to the more recent damage; however, comprehensive research in historic archives (i.e., Byzantine) would be worthwhile.
The subsurface erosion and/or internal and contact suffusion, especially of isolated sand lenses above the gravel sands of the Lower Terrace, are considered a feasible explanation for nine of the 11 damage areas (Fig. 20), resulting in p = 0.82. The seepage models show that for a punctured flood loam layer, the hydraulic conductivity of the subsurface is high enough to quickly transport large amounts of water towards the sand layers during rainstorm events. Large parts of the study area were hydraulically sealed; this applies to the Praetorium itself but also to the current town hall square. During the imperial period, a sophisticated and well-maintained dewatering system existed in Cologne. In addition to rainwater accumulation, the water from the Eifel aqueduct with a maximum daily discharge of 21,000 m3 (Brinker 1986) required drainage towards the Rhine River front and ultimately into the river. Due to the fast progressing sedimentation of the northern sewer in the fourth century CE (Schütte and Gechter 2011), the sediments, which even included several dog skulls, are a clear indication of a lack of maintenance and a rapid decline of the dewatering system. Several water channels existed under the foundations of the Praetorium. A major drainage canal conducted rainwater from a collecting basin northwest of the Octagon, part of the CVM, towards the area in front of the Octagon, where it met with other canals in front of an outlet through the Praetorium front façade. If these and other drainage systems were no longer maintained after the Roman period, as with the northern sewer, uncontrolled seepage was the likely cause of damage.
The massive foundations of the front wall of the Praetorium directly east of the Octagon reaching a depth of 42.7 m asl (which is 3 m deeper than in the western part of the Octagon, (Precht 1973)) and consisting of opus caementicium show that the Roman engineers were aware of stability problems in this area. This interpretation is supported by several retaining walls from earlier periods within the terrace slope. The change in the foundation toward the north to a much more simple pile grading foundation has so far been interpreted as a consequence of two building phases. However, an additional reason could be the original topography and that the engineers adjusted the strength of the foundations to accommodate the geological conditions of the subsurface, for example, in form of sand lenses. The deformation analysis based on the CVM indicates a deficiency of ∼20 m3 in the subsurface east of the Octagon foundation. This is roughly equivalent to a 0.5 m thick sand layer stretching over an area of 5 × 8 m, and it is plausible that such an amount of material was washed out. Taking into account that the grain size of the analyzed sand lenses from the outcrops in the excavation are highly sensitive to subsurface transport (Prinz and Strauss 2006) and that given the climate conditions, heavy rainstorms, especially during summer season would not be unusual, repeated internal suffusion, contact suffusion, and finally erosion becomes a feasible cause for damage.
The collapsed hypocaust floor under the synagogue shows an inclination of 200 mm/m (1:5) at its base where it is sitting on top of an isolated sand lens. State-of-the-art engineering knowledge (i.e., Prinz and Strauss 2006) predicts substantial fractures in walls and structural damage at inclinations of 1:150 and total loss of buildings at 1:20, four times less than our observation. The convex bend of the subsurface (Fig. 7) introduced significant tensile forces in the concrete floor on top of the tile columns, leading to a collapse of the structure. This makes a subsidence due to sediment loss a feasible explanation of the hypocaust collapse.
The structural damage of the structure underneath the women’s synagogue including the extrusion of a part of the massive foundation together with its 3.3° inclination in the lower part is compatible with a washout of sediments in the subsurface in the northeastern corner of the women’s synagogue. The women’s synagogue is only 5 m from the Roman well that showed even recent sediment bulging during the excavation. Sharp-edged fractures cutting through blocks of the upper Medieval section are an indicator of a sudden event. From the archeological context, it is not yet clear whether the lower Roman and upper (possibly) Medieval sections were damaged at the same time. If the damage occurred concurrently a vibration-induced damage is not likely; vibrations strong enough to break and shift parts of the strong Roman foundation, would have severely damaged or even collapsed the rather fragile Medieval top construction (Fig. 8).
The damage rated as not feasibly produced by seepage (Fig. 20) occurs in the northern sewer. A possible explanation for the mobilized sediments is the repeated Rhine River flooding and the structural damage, especially the cracked roof of the eastern sewer section, is offered by the dynamic impact of WWII bombing.
Hinzen and Schütte (2003) estimated liquefaction potential in the uppermost 3 m of the Praetorium site with factors of safety of 1.02 based on a simple level-ground liquefaction analysis. However, our updated subsurface model and the ground water table data indicate that such shallow sections stay dry even during high water levels of the Rhine River. The groundwater-saturated layers are more than 10 m below the surface, and due to the large overburden are unlikely to liquefy, an interpretation, which is also supported by our dynamic slope stability calculations.
Based on the data acquired in the frame of this study (including the comprehensive virtual model and the subsurface model) and the current status of the excavation, a seismogenic cause of all the damages is less probable than previously hypothesized. In particular, the results of the slope stability tests indicate a subsurface stable even under significant dynamic loading. Barring new data with a clear suggestion of earthquake damage from the planned excavations of the southern part of the Praetorium, we favor the interpretation of seepage effects and subsurface erosion as explanation for the settlement-induced building failures in more than one event and WWII bombing in case of the broken sewer roof. Subsurface mobility of sandy material is also discussed as a possible cause of the dramatic collapse of the historic Cologne city archive in 2009, which was located next to a subway construction site on the Lower Terrace of the Rhine River.
We are grateful to the whole team the Archaeological Zone Cologne for the support of the fieldwork especially to W. Münter. We thank M. Gechter, K. Kliemann, S. Schütte, and M. Wiehen for numerous discussions on all aspects of the excavation. S. Oesinghaus of Kühn Geoconsult significantly supported the determination of geotechnical parameters and R. Esser helped with the field and laboratory work. M. Wegner worked on the seismotectonic model and the synthetic groundmotions. G. Schweppe and H. Kehmeier helped during fieldwork, with the creation of some figures and text processing, respectively. The Kölner Verkehrsbetriebe provided information on exploratory boreholes for the subsurface model. The Stadtentwässerungsbetriebe Köln provided the data on Cologne precipitation. We thank S. Marco, T. Niemi, T. Tuttle, K. Vanneste, and K. Verbeeck for their discussions on mobilized sediments. Comments from an anonymous reviewer were of great help and we appreciate the work by editorial manager D. Javinal and chief editor T. Dahm. This work was financed by Deutsche Forschungsgemeinschaft (DFG) grant HI 660/2-1.
This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.