Landslide impact on the archaeological site of Mitla, Oaxaca

Some large landslide deposits are recognized within and around the Mitla archaeological site, which is located in a very seismically active region cut by the Tehuacan‐Oaxaca fault system. Previous studies have shown that this fault system has produced coseismic ruptures and associated earthquakes in the past. Our work on the Mitla landslide deposits clearly demonstrates that their morphology and nature are typical of dry‐rock avalanches. Thus, collapses of rock already weakened by alteration and/or weathering, and unstably mobilized by intense rainfall, are discarded as possible triggers of the avalanche events. We instead propose that the landslides were triggered by earthquake activity. Moreover, our data indicate that part of the original Mitla settlement lies buried under a large rock‐avalanche deposit produced sometime during the post‐Classic period (900–1520 AD). At that time, the city of Mitla was inhabited by over 10,000 people as estimated from archaeological reconstructions. This devastating landslide almost entirely obliterated Mitla, which at the time of the Spanish arrival still existed as a city, but much reduced in area and populations.

landslides stands at less than 1 km from the main entrance to the Mitla ceremonial center. The landslide morphology and close proximity to Mitla lead us to ask if this event perhaps occurred during the occupation of Mitla by the Zapotec Empire from 900 to 1521 AD.
Although the Mitla archaeological site stands only a few hundred meters from the landslide front, surprisingly, this deposit has not been previously described in the geological literature (Ferrusquía-Villafranca, 1990;Ferrusquía-Villafranca, Wilson, Denison, McDowell, & Solorio-Munguía, 1974;Iriondo, Kunk, Winick, & C.R.M., 2004) nor in the archaeological studies (Robles-García, 2016). In this study, we present the results of the morphological and geological characterization of the landslide deposits and their spatial and temporal distribution with respect to the pre-Hispanic ceremonial center of Mitla. In particular, we focused on the landslide depositional chronology, the possible triggers, and its' possible repercussions for the pre-Columbian population of Mitla, the second most important human settlement of the Oaxaca valleys. By combining the geological and shallow geophysical techniques (e.g., electrical tomography), we provide new evidence that suggests that the ceremonial center could have been directly affected by the landslide.

| Tectonic setting of Oaxaca
The state of Oaxaca is one of the most active seismic zones in southern Mexico because of the subduction of the Cocos Plate beneath the North American Plate along the Middle American Trench (Pardo & Suárez, 1995) (Figure 1a). This tectonic environment produces both earthquakes along the subduction zone and intraplate shallow earthquakes (Figure 1b). The earthquake of September 7, 2017, in Oaxaca, and the shallow Xalapa earthquake that took place in 1920 (Ms 6.4, Córdoba-Montiel, Krishna Singh, Iglesias, Pérez-Campos, & Sieron, 2018) are examples of these two types of seismic activity. Mitla is located within an area that is subjected to both subduction and intraplate earthquakes, the latter is associated with movements along the boundary of two tectonic terranes exposed in the Central and Tehuacán Valleys. In fact, the Central Valleys of Oaxaca are part of the Oaxaca fault system that forms the tectonic boundary between the Zapoteco and Cuicateco terranes (Figure 1c).

| The Mitla archaeological site
Mitla is located south of the City of Oaxaca, capital of the State of the same name, along the NW-SE Etla Valley (Figure 2). Mitla's climate is dry temperate (Koppen-Geiger BsK), with an average rainfall of 623 mm and an average temperature of 17.4°C. The name Mitla or Mictlán is of Náhuatl origin and means "Place of the Dead" or "Underworld." In Zapotec, it is called Lyobaa which means "Place of Burials"; in the Mexica language, it remained Mitlán, "Place of the Dead," or "place of many corpses," and was Hispanicized to Mitla (Corner, 1899). Mitla's zenith occurred between 950 and 1521 AD, when it covered an estimated area of over 7,000 km 2 , with a population of approximately 10,000 inhabitants. Among Mitla's specific attractions are buildings decorated with elaborate mosaic fretwork (grecas) showing variations of the same geometric design, and cross-shaped tombs that have been found beneath the palaces, in which important people and priests were presumably buried. The archaeological complex includes several F I G U R E 1 (a) Tectonic sketch of southern Mexico showing the tectonic terrains described by Campa and Coney (1983) and the tectonic plate configuration. (b) Seismicity reported (1917)(1918)(1919) with depths between 0 and 20 km deep by the Servicio Sismológico Nacional, greater earthquake (stars) and focal mechanisms from IRIS consortium.

| Local geology
The stratigraphic succession around Mitla is mainly composed of the basement rocks and the rhyolitic ignimbrites that form the Sierra La Calavera ( Figure 4). On top of this stratigraphic succession, we have faulted fluvio-lacustrine deposits comprising layers of conglomerates, sands, and clays that contain large Pleistocene vertebrates (Ferrusquía-Villafranca, 1990 andFerrusquía-Villafranca et al., 1974). All these rocks are transected by NNW-SSE left-lateral strike-slip faults. The main fault here called the Calavera, is one of the regional faults that bound the Zapoteco and Cuicateco terranes.
The stratigraphy of the area is dominated by the rhyolitic ignimbrites that are very well exposed on the hanging wall and scars of the Calavera fault ( Figure 4). This ignimbrite was first described by Williams and Heizer (1965)  Overlying the rock-avalanche deposit are deposits that vary from debris flows to hyper-concentrated flows, comprising the same ignimbrite blocks embedded in a silty-clayey matrix. Above these deposits, we found a layer of mudflow deposits that contain ceramic shards ( Figure 5d).
We calculated the area and volume of the rock avalanches using ArcMap 10.2 software and the DEM ( Figure 6). The rock avalanches cover an area of 4.19 km 2 and possess an average thickness of 0.06 km.
Therefore, we estimate a minimum total volume of approximately 0.2 km 3 for the two deposits combined with about 0.1 km 3 for each.

| Coefficient of friction (H/L)
The vertical drop versus the maximum runout distance traveled by a landslide is known as the apparent friction coefficient (H/L) that was introduced by Hsü (1975). The coefficient of friction was then used to describe debris avalanches in volcanic terrains (Ui, 1983), and then by Francis (1993) for various types of landslides (Legros, 2002;Morelli et al., 2010Morelli et al., , 2016. For small-size landslides, the apparent coefficient of friction is about 0.6 while for larger size landslides, this value typically is about 0.2. For the case of the Mitla landslide, we obtained a maximum drop of 640 m with a maximum runout of 1,633 m F I G U R E 5 Different aspects of the Mitla rock-avalanche deposit. (a) A block exposed on the surface of the rock avalanche that is approximately 6 m in diameter. (b) Reverse faults in the ignimbrite rocks exposed on the hanging wall of the scar, (c) Quarry exposing two main beds of the rock-avalanche deposit. (d) Lithology exposed at the rock-avalanche front. The tip part of the distal avalanche covers part of the archaeological site with at least 2 m of debris [Color figure can be viewed at wileyonlinelibrary.com] that yielded an H/L = 0.39 (Figure 7). This value in the chart of Figure 7 correlates with dry debris avalanches of nonvolcanic origin (Dade & Huppert, 1994). The Mitla debris-avalanche deposit comprises shattered blocks supported by a coarse ash matrix with no other associated pyroclastic deposits and with no evidence of transport by water (e.g., voids, rounded blocks). Therefore, it is plausible that the triggering mechanism for the landslide could have been an earthquake that promoted the downhill remobilization of the rocks. Hsü (1975) introduced an indicator regarding the mobility of landslides called excessive travel distance. This parameter (Le) corresponds to the horizontal distance traveled in excess by the event (i.e., above that expected) for a dry rigid mass sliding down on a slope with a normal coefficient friction of tan 32. This value is small in comparison with those calculated for rock avalanches that involve water among the triggering factors.

| Geoslope modeling
In characterizing the conditions that can trigger slope movement, it is necessary to understand and quantify the main forces involved, namely shear strength and shear stress. To perform such slope-stability analysis, for this study we used the GeoSlope computer program (GEO-SLOPE International Ltd, 2008a), in particular its software package SLOPE/W. In assessing if the slope will move or not, GeoSlope calculates an indicator called "safety factor" (or sometimes, but less commonly, "security factor").
This factor is an expression of the relation between the forces resisting movement (that must be sufficiently strong to prevent slope movement) and the gravitational forces (that would cause the slope to fail). If the safety-factor value is equal to 1, the slope is considered at the limit of equilibrium conditions. Therefore, values of 1 or higher means that the slope is stable, and values lower than 1 means that the slope is unstable (Duncan, Wright, & Brandon, 2014). We calculated the safety factor in a distributed way over the entire investigated section (yellow line in Figure 9 inset). The results show that the upper parts of the slope, from 350 to 1,300 m, and between 450 and 650 m, the safety factor is lower than 1 (in red or orange in Figure 9 profile), which means it is unstable.
This portion of the slope is near the crown of the large break caused by the La Calavera fault. The profile also indicates that farther downslope the safety factor increases (in green or blue) to values higher than 1, indicating that this portion of the slope is stable. These results allowed us to establish the boundary of the slope and to detect where the movement was resisted (Fidolini, Pazzi, Frodella, Morelli, & Fanti, 2015).

| Application of quake/W
Slope-stability analysis can be studied both under static (see Section 6.2) and/or dynamic conditions (GEO-SLOPE International Ltd, 2008b). The analysis for dynamic slope behavior is performed when the landslide triggering mechanism is an earthquake, and can be carried out by means of a different GeoSlope software package, called Quake/W (Garevski, Zugic, & Sesov, 2013). The input data are the slope-boundary conditions and an accelerogram of the earthquake (Figure 10) Crozier (1992) proposed six criteria to link landslides to a seismic origin in New Zealand that are given below: (a) The occurrence of modern seismic activity in the study region that has triggered land- that cannot be described solely on the basis of geological and geomorphic conditions. In a recent study of landslides in Asia, Strom and Abdrakmatov (2018) considered that with the exception of criteria 3 and 4 of Crozier (1992), most of the other criteria appeared to be met in their compilation, as we also do observe for the Mitla landslides in Oaxaca. Taking into account the geomorphology of the Mitla rock avalanche (e.g., steep front) and the physical structure of the deposit (composed of shattered blocks with an intraclast matrix produced by F I G U R E 9 Results obtained with the static stability analysis module of GeoSlope model along a longitudinal section located in the center of the landslide (as highlighted by the yellow line on the 3D model). The variation of the static safety factor (stable area with a factor higher than 1 in green and blue and unstable area with a factor lower than 1 in yellow and red) is shown from the output model (a) in correlation with the reconstruction of the material distribution inside the landslide mass (b) [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 1 0 Record of the earthquake in the accelerogram used to study Mitla as an analog [Color figure can be viewed at wileyonlinelibrary.com] attrition during transport), we suggest that the deposit is a typical dry rock avalanche. In fact, the coefficient of friction (H/L = 0.39) also suggests a dry origin for the landslide (Dade & Huppert, 1994). The internal aspect of the rock-avalanche deposit does not reveal subrounded to rounded blocks nor voids in the fine-sized matrix that would support the presence of water during transport typical of debris flows. Another important aspect of the avalanche deposit is that it comprises fresh ignimbrite blocks and, therefore, no intense alteration or weathering of the rocks could have favored the collapse. Taken collectively, these features strongly support the argument that the collapse was not triggered by rainfall. Therefore, it is quite probable that the cause of the rock avalanche may be of seismogenic origin.
Earthquakes are one of the most common triggers of large-scale bedrock landslides, as documented worldwide (Strom & Abdrakmatov, 2018). Earthquakes and landslides are often linked, because earthquakes cause fracturing and destabilization of solid rocks weakened by their rheology, their strata geometry, or water saturation, and, occasionally, also entail the remobilization of rocks (Bommer & Rodríguez, 2002;Rodríguez, Bommer, & Chandler, 1999).
Earthquakes of moderate or high magnitude can trigger landslides that sometimes cause human deaths and significant infrastructure damage. It has also been established that massive landslides can be triggered by the combination of a shift in pore pressure caused by rain or a mild earthquake. Keefer (1984Keefer ( , 1994 concluded that earthquakes can cause different mass movements in different geological environments and that the geology influences each movement in a different manner. More vulnerable environments are usually highly fractured, weathered, sheared, or soft rocks. Volcanic ashes, noncohesive residual soils, and alluvial and colluvial deposits are also susceptible. Keefer (1994) developed a chart to assess the relationship between earthquakes and the volume of the sediments produced by them. Even though it was difficult to identify the exact number of landslides per earthquake, Keefer (1984Keefer ( ,1994 showed

| Earthquakes and archaeology
Earthquakes and archaeology may, at first glance, appear as two entirely different fields of research; however, in Mesoamerica, as in Mesopotamia, Egypt, Greece, and Anatolia, archaeological remains provide evidence of seismic activity during ancient human history.
García Acosta and Suárez-Reynoso (1996) assessed the impact of seismic events on pre-Columbian cultures in México for the first time.
During the past 20 years, archaeoseismology (Giner-Robles, Source of epicenter data S. K. Singh (IG) F I G U R E 1 1 Calculation of the probable magnitude of a possible earthquake that could have triggered the Mitla's landslides (following Keefer, 1984). The red arrow corresponds to the Mitla rock avalanche and the green arrow to the NW Mitla rock avalanche [Color figure can be viewed at wileyonlinelibrary.com] Rodríguez-Pascua, Silva, & Pérez-López, 2018;Stiros & Jones, 1996) has introduced new approaches to study the effects of past earthquakes and, therefore, their impacts on ancient cultures. Earthquakes have been and remain frequent in the history of Mesoamerican cultures. In the Zapotec language of the Oaxaca region, they were called xòo, and in the Nahuatl language of central Mexico, ollin (movement).
One of the field-excursion books of the 1906 International Geological Congress in Mexico referred to a Mitla earthquake that occurred in 1495 by using the symbols of the Codex Mendoza. More recently, Garduño-Monroy (2016) described 12 seismic events recorded with a tlalollin in the codex Telleriano-Remensis. The most powerful of them occurred in 1507 ("año dos cañas") with a VIII intensity (Figure 12), causing an avalanche of water, mud, and stones (debris flow) that killed more than 2,000 warriors in the River Tuzac. A modern analog of this event is the 5.4-magnitude earthquake that occurred on June 6, 1994, southwest of Nevado del Huila volcano, Colombia (Ávila et al., 1995). This earthquake produced thousands of landslides that poured into the Paez River, producing a massive wave of mud that traveled more than 120 km downstream and killed 274 people (Scott, Macias, Naranjo, Rodriguez, & McGeehin, 2001).
During his studies in the Mitla area, Beals (1933)  from the partial destruction of highly esteemed, pre-existing buildings.
Was the cause of this destruction related to some natural disaster, such as perhaps a landslide? Archaeologists report that, during its prime, Mitla had a population of about 10,000 people, a figure equal to the current population of the present-day city, which covers an area over 1.5 km 2 .
However, if we consider the five remaining pre-Columbian buildings, the total area of ancient Mitla at the time was 0.18 km 2 , which is too small of an area to hold 10,000 people, thereby further suggesting that other parts of the ancient city are also buried by the rock ava-

DATA AVAILABILITY STATEMENT
Data available on request from the authors.The data that support the findings of this study are available from the corresponding authors upon reasonable request.