Vulnerability and Site Effects in Earthquake Disasters in Armenia (Colombia). II – Observed Damages and Vulnerability

54 55 Damage in Armenia, Colombia, for the 1999 (Mw6.2) event was disproportionate. We analyse the damage report as a 56 function of number of storeys and construction age. We recovered two vulnerability evaluations made in Armenia in 1993 57 and in 2004. We compare the results of the 1993 evaluation with damages observed in 1999 and show that the vulnerability 58 evaluation made in 1993 could have predicted the relative frequency of damage observed in 1999. Our results show that 59 vulnerability of the building stock was the major factor behind damage observed in 1999. Moreover, it showed no significant reduction between 1999 and 2004. small downtown district of the city where the two data sets overlap. The comparison of the two vulnerability studies, in 1993 and in 2004, allows an assessment of the changes in vulnerability in the city as a consequence of a destructive earthquake, even if the method used was different and the studied zones overlap only partially. We show that building vulnerability was the main factor behind the heavy damage toll in Armenia during the 1999 earthquake. Our results substantiate the improvement of engineering practice with time and provide evidence of the efficacy of simple methods to evaluate vulnerability. However, they also strike an alarm bell as they show that vulnerability in Armenia remains high. Our results offer an unusually complete analysis of the major factors behind seismic risk in a typical medium size city in Colombia. Seismic risk mitigation in Armenia, and 111 in similar midsize cities in Latin America, requires an increase in the number of permanent seismic stations and support of additional efforts to improve our understanding of moderate size seismic events. them 286 with arbitrary weights based on expert opinions to compute a vulnerability index for each building in the sample. The 287 weights used to estimate a vulnerability index had to be modified from those used in 1993 given that less information on 288 each structure was available. The VI results for the 2004 study may thus have a constant bias. We could assign a 289 vulnerability index to 1,217 buildings, out of the 2,525 counted in 2004. The building categories that could be identify 290 between the two studies were bahareque, unconfined masonry and frame structures. VI values were grouped in three 291 categories: low (VI between 0 and 20), medium (VI between 20 and 40), and high (VI larger than 40). The results in Figure 8 show that the relative proportions are maintained between 1993 and 2004: most buildings in that sector have still 293 high vulnerability in 2004 and less than 20% have low VI. Our results suggest that significant improvements in the relative 294 vulnerability occurred in the 11-year period between 1993 and 2004. High vulnerabilities are still predominant in 295 downtown Armenia, in spite of the destruction of weak buildings in the 1999 earthquake and the reinforcement carried 296 out during the reconstruction of the city. It may be hoped that this result will prompt local authorities to take decisive 297 actions to mitigate seismic risk in Armenia. A starting point could be to replicate the use of simplified procedures to 298 estimate vulnerability to evaluate possible changes in the 16-year period since 2004.


68
significant events. The first building code in the country was published in 1984 (CCCSR-84, 1984), partly as a result of 69 the heavy toll caused by the Popayán earthquake in March, 1983(Ingeominas, 1986. Increasing building requirements 70 have improved earthquake resistance, for example phasing out non engineered construction. The development of 71 earthquake engineering has led to a decrease in the vulnerability of buildings in Colombia but progress has been slow, in 72 pace with the development of building codes. In addition, as favoured construction styles evolve, additional challenges 73 appear. For example, the cost of land pushes current housing projects consisting of tall concrete structures for which there 74 is little experience regarding their seismic behaviour in that country. Instrumenting some of those buildings to analyze 75 their motion during small earthquakes would provide useful data and may eventually become a necessity (e.g., Meli et 76 al., 1998). Meanwhile, it is important to learn as much as possible from past destructive events.

78
Damage evaluation after large earthquakes is recognized as a primary input to understand structural response subject to 79 dynamic excitations. It offers valuable data on the behaviour of structures to actual seismic motion. In addition to very 80 significant efforts like GEER (Geotechnical Extreme Events Reconnaissance, 2020), local initiatives have contributed 81 significantly to understand damage occurrence, especially in relation to site effects (e.g., Montalva et al., 2016;Fernández 82 et al., 2019).

84
One seismic event that has had a long lasting impact in Colombia is the January 25, 1999, earthquake in the Quindío 85 department, close (18 km) to the city of Armenia. This relatively small (Mw6.2), normal fault earthquake had profound 86 economic and social consequences in the country. There was only one accelerograph in Armenia, and it recorded PGA of 87 518/580/448 gal in the EW/NS/Z components. Strong ground motion duration was very short (smaller than 5 s) and 88 ground motion energy peaked at periods shorter than 0.5 s. The source of the main shock and aftershocks was studied in 89 Monsalve-Jaramillo and Vargas-Jiménez (2002), while macroseismic observations were presented in Cardona (1999).

90
The city of Armenia sustained heavy damage (maximum intensity was IX in EMS-96 scale): 2000 casualties and 10,000 91 injuries due to the collapse of 15,000 houses, with a further 20,000 houses severely damaged (SIQ, 2002). Site effect 92 evaluation during this event in Armenia was addressed by Chávez-García et al. (2018). Earthquake and ambient noise 2 Colombian Building Codes and Practice Evolution

117
This paper will obviate a discussion of the geological setting of Armenia, as it can be found in Chávez-García et al. (2018).

118
The coffee growing region was occupied during the second half of the 19th century. For this reason, data on historical 119 earthquakes is scarce, even though it is located in a zone of high seismic hazard (the current Colombian building code 120 prescribes a PGA of 0.25 g for Armenia for a return period of 475 yr). During the 20 th century, about eight earthquakes 121 occurred in the region producing intensities as large as IX (Espinosa, 2011  In the aftermath of the 1999 event, the Sociedad de Ingenieros del Quindío (Quindian Society of Engineers) organised 152 teams that made a detailed evaluation of damaged structures in Armenia (SIQ, 2002). The status of a building is 153 determined by the attributes of damage level, damage type and usage status (Tang, et al., 2020). The priority was to 154 distinguish between those buildings that did not pose a risk to occupants from those that must be evicted. The template 155 used to qualify buildings allowed to grade the damage sustained by buildings and included information on year of 156 construction, structural system, and number of storeys. SIQ (2002) classified observed damage using a colour scale: 157 • Grey. Very light or no damage at all.

158
• Green. The building can continue to be used. Although some damage is apparent in non-structural elements, it 159 poses no risk to occupants 160 • Yellow. Significant damage, to the point that partial occupancy restriction is required. The structure is not 161 evaluated as unsafe but access to parts of it must be restricted.

162
• Orange. Unusable structure. Damage to the structure implies a high risk and the building cannot be occupied.

163
• Red. Total collapse or danger of collapse due to severe damage to the structure or its foundation.

164
This scale is quite standard and very similar to that proposed by the European Seismological Commission (Xin et al., 165 2020). For our purpose, we have simplified this scale. We use light damage to refer to structures classified in grey or 166 green. Moderate damage in this paper is used for buildings classified as yellow. Finally, severe damage corresponds to 167 structures classified as orange or red. The SIQ (2002) report presents an inventory of 43,023 structures classified as a 168 function of damage sustained. From this total, data for 1,946 sites could not be used due to incomplete information that 169 made it impossible to locate them on a map. This number suggests a lower limit for the uncertainties in our database,

175
Five categories were used to classify the buildings structuring type, following CCCSR-84 (1984). In order of decreasing 176 seismic performance, the first four categories are: frame structures, confined masonry, unreinforced masonry, and (2002), is "none of the preceding", named as "other" in the following. This last category was used to refer to buildings 179 using hybrid structuring systems, a mix of different materials, and unstructured houses mixing wood with other elements.

244
In 1993, different sectors of the city were sampled but not all of the data were preserved. We analyse the results for the 245 downtown sector presented in López et al. (1993), shown in Figure 2. A census was made to count the number of structures 246 of each type. In the downtown sector, 3,364 buildings were counted and assigned to one of three categories: bahareque 247 structures (908), unreinforced masonry structures (1,877), and frame structures (579). It was not possible to evaluate, 248 even in a simplified way, all those structures. For this reason, a small sample of 84 buildings was designed, assuming 249 normal distribution and choosing a 95% confidence level of the extrapolation of the results to the total population. The 250 84 buildings were randomly selected in the field and the vulnerability of each of them was evaluated using the procedure 251 described in Tassios (1989), which is very similar to that described in Inel et al. (2008) Figure 2). Percentages for VI values were extrapolated from the numbers 262 determined for the 84 building sample. In this figure, we counted together moderate and severe damage, while VI was 263 classified in two groups: larger and smaller than 20. We observe a very good correlation between VI estimated in 1993 264 and damage observed during the 1999 earthquake, six years later. Thus, the approximate procedure used to estimate VI 265 in 1993 was effective to predict dynamic behaviour during that earthquake.

267
In addition to comparing extrapolated VI with damages for the downtown district, we may ask another question. How did 268 each one of the 84 buildings, whose VI was evaluated, fare during the 1999 earthquake? This question has no simple 269 answer due to different georeferencing systems for the two surveys (vulnerability and damage) and incomplete data. Only 270 28 out of the 84 could be confidently identified. The unidentified buildings could be absent from the damaged buildings 271 database because they suffered no damage or because their recorded location was inaccurate. Figure 7 shows a whisker 272 plot of the observed VI values against observed damage for the 28 buildings that could be identified in both databases.

273
VI values are well correlated with observed damage. Figure 7 shows that severe damage may be associated with an 274 average VI of 44, moderate damage with an average VI of 32, while light damage corresponds to an average VI of 16.

276
Consider finally the vulnerability study made in 2004 (Cano-Saldaña et al., 2005). The procedure used was very different 277 and followed that of Velásquez and Jaramillo (1993). Cano-Saldaña et al. (2005) computed expected losses for three 278 different events, considered to pose the largest seismic hazard for Armenia. A required input for them was an estimate of 279 the vulnerability for the building stock, and this is the data we recuperated from that study. Cano-Saldaña et al. (2005) 280 selected a sector of the downtown district that overlaps only partially with the district sampled in 1993. It is shown with 281 dot-dashed line in Figure 2. They tallied every building in that sector, a total of 2,525 land plots. For each one of them, a 282 template simpler than that of 1993 was completed including data on structuring type, number of storeys, roofing type, Armenia and forced important improvements in engineering practice. The large economic consequences led the 305 government to add a new tax to pay for reconstruction: a levy of 2‰ was imposed on every bank transaction in the 306 country. Earthquake disasters occur rarely and therefore seismic risk is seldom a priority. In Armenia region, the first two 307 accelerographs were installed in 1994: in the campus of Universidad del Quindío, and in Calarcá (a neighbouring town, 308 10 km to the SE of Armenia). To date, they continue to be the only accelerographs in operation. As mentioned above, the 309 mandatory microzonation study of Armenia is still due.    Ingeominas, 1999.] 498 499 500 501 502 503 https://doi.org/10.5194/nhess-2020-385 Preprint. Discussion started: 21 December 2020 c Author(s) 2020. CC BY 4.0 License.

505 506
Figura 2 Contornos de la profundidad de la interfaz (en m) en la base de los depósitos de ceniza que cubren la ciudad de Armenia. Los 507 símbolos muestran la ubicación de los 36 sondeos verticales eléctricos donde se midió la profundidad de esa interfaz. La gruesa línea 508 sólida que cruza la ciudad de norte a sur indica el rastro de la falla de Armenia. El polígono de línea sólida dentro de la ciudad muestra 509 la extensión del distrito del centro cubierto en el estudio de vulnerabilidad de 1993. El esquema de línea pequeña y salpicada de puntos 510 muestra el área cubierta por el estudio de vulnerabilidad realizado en 2004. [Modificado de Ingeominas, 1999.]   the given structuring type and shows the relative incidence of light, moderate, and severe damage as a function of the 560 time period where the structure was built (before 1959, between 1960 and 1984, between 1985 and 1997, and later than 561 1998). The data shown corresponds to the downtown district whose outline is shown in Figure