Earth and Planetary Science Letters 205 (2002) 81^89 www.elsevier.com/locate/epsl
Precursory di¡use carbon dioxide degassing signature related to a 5.1 magnitude earthquake in El Salvador, Central America J.M.L. Salazar a; , N.M. Pe¤rez a , P.A. Herna¤ndez a , T. Soriano b , F. Barahona b , R. Olmos b , R. Cartagena b , D.L. Lo¤pez c , R.N. Lima a , G. Melia¤n a , I. Galindo a , E. Padro¤n a , H. Sumino d , K. Notsu d a
d
Environmental Research Division, Instituto Tecnolo¤gico y de Energ|¤as Renovables (ITER), 38611 Granadilla de Abona, S/C de Tenerife, Canary Islands, Spain b Instituto de Ciencias de la Tierra, Universidad de El Salvador, San Salvador, El Salvador c Department of Geological Sciences, 316 Clippinger Laboratories, Ohio University, Athens, OH 45701, USA Laboratory for Earthquake Chemistry, Graduate School of Science, University of Tokyo, Hongo, Bunkyo-Ku 113-0033, Tokyo, Japan Received 30 January 2002; accepted 5 August 2002
Abstract Anomalous changes in the diffuse emission of carbon dioxide have been observed before some of the aftershocks of the 13 February 2001 El Salvador earthquake (magnitude 6.6). A significant increase in soil CO2 efflux was detected 8 days before a 5.1 magnitude earthquake on 8 May 2001 25 km away from the observation site. In addition, pre- and co-seismic CO2 efflux variations have also been observed related to the onset of a seismic swarm beneath San Vicente volcano on May 2001. Strain changes and/or fluid pressure fluctuations prior to earthquakes in the crust are hypothesized to be responsible for the observed variations in gas efflux at the surface environment of San Vicente volcano. > 2002 Elsevier Science B.V. All rights reserved. Keywords: precursory signatures; carbon dioxide; difusse degassing; earthquake; San Vicente volcano; El Salvador
1. Introduction Utilizing geochemical parameters to forecast earthquakes is a di⁄cult task, in part because evidence of precursory geochemical events is di⁄-
* Corresponding author. Tel.: +34-922-391000; Fax: +34-922-391001. E-mail address:
[email protected] (J.M.L. Salazar).
cult to acquire. Impediments to detecting precursory events related to impending earthquakes include a scarcity of the number of monitoring sites, lack of long-term data records, and ine¡ective integration of multidisciplinary data (e.g. geophysics and geochemistry). Additionally, the existence of geochemical precursors and/or the ability to ascribe particular geochemical events to particular seismic events is regarded with skepticism and not universally accepted [1,2]. Despite such unfavor-
0012-821X / 02 / $ ^ see front matter > 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 1 2 - 8 2 1 X ( 0 2 ) 0 1 0 1 4 - 2
EPSL 6456 28-11-02
82
J.M.L. Salazar et al. / Earth and Planetary Science Letters 205 (2002) 81^89
able circumstances and although geochemical and hydrological observations preceding seismic behavior are di⁄cult to explain or model, recent work suggests that £uid physical and chemical behavior is clearly sensitive to crustal stress changes [3^6] ; however, geochemical and hydrological observations, acting as earthquake precursors, should be considered as strain indicators and not as strain meters [7]. On 13 January 2001, a 7.6 magnitude earthquake (12.8‡N, 88.8‡W) occurred o¡ the El Salvador coastline within the subduction zone of the Cocos plate (60 km depth), causing extensive damage throughout the entire country [8]. By 19 January 2001, about 660 aftershocks had been registered. On 13 February 2001, a 6.6 magnitude earthquake with an epicenter about 20 km west of
the San Vicente volcano (Fig. 1) damaged and destroyed numerous towns and villages in the north area of this volcano, causing many fatalities. At least seven major seismic events have been registered in the surrounding inhabited areas of the San Vicente volcano since 1783, as well as six seismic swarms in the 90s (Centro de Investigaciones Geote¤cnicas, El Salvador). In March 1999, a seismic swarm comprising 934 events in 23 days was registered north of the San Vicente volcano. Focal mechanisms for 10 events showed a strike-slip behavior with only a small normal component and a fault plane of E^W orientation, coincident with the local fault system within the epicentral zone [9]. Seismicity in the region was intense after the 13 January and 13 February 2001 earthquakes. It declined from February to
Fig. 1. Map of the San Vicente volcano and surrounding areas in El Salvador, with inset showing the Central American Volcanic Belt, San Vicente volcano, and the earthquakes of January and February 2001. The star shows the location of the automatic geochemical station (GS) for the continuous monitoring of di¡use CO2 e¥ux at the San Vicente volcano.
EPSL 6456 28-11-02
J.M.L. Salazar et al. / Earth and Planetary Science Letters 205 (2002) 81^89
March with only a few small-magnitude events around 9 April. In May 2001, a new seismic swarm was registered beneath the northeastern £ank of the San Vicente volcano (Fig. 1) at depths down to 18.4 km. A total of 175 seismic events with a magnitude of between 2.1 and 5.1. were recorded in 17 days. The San Vicente volcano (2180 m a.s.l.) is a Quaternary composite andesitic volcano located 50 km east of San Salvador city and has been inactive for the last 3000 years. Its paired edi¢ce rises from the Central American Graben, an extensional structure parallel to the Paci¢c coast. The northern sector of the Central American Graben exhibits a system of E^W trending faults with displacements of up to 300 m [10]. Fumaroles (98.2‡C) and hot spring waters are present along radial faults at two localities on the northern slope of the volcano (Aguas Agrias and El In¢ernillo). Aiuppa et al. [11] hypothesized that these discharges are linked to a 250‡C hydrothermal system at a depth of 1200 m with an estimated CO2 partial pressure of 14 bar. Carbon dioxide has been used as a tracer of sub-surface magma-degassing due to the fact that it is the major gas species after water vapor in both volcanic hydrothermal £uids and magmas [12]. Degassing from magma and hydrothermal aquifers creates gas anomalies at the surface as a result of a combination of advective and di¡use transport mechanisms [13]. Relatively high CO2 £uxes correlate with both high heat £ux areas (related to active and dormant volcanism) and areas with deep fractures/faults which transport CO2 derived from magmatic and/or decarbonation processes [14]. Irwin and Barnes [15] have shown that the global distribution of CO2 degassing areas correlates with zones of seismicity and high tectonic stress. CO2 degassing is also commonly related to active seismic faults characterized by high permeability and porosity acting as drains in the crust.
2. Methods With the aim of detecting changes in di¡use CO2 degassing related to seismic activity at the
83
San Vicente volcano, an automatic geochemical station (WEST Systems, Italy) was installed on 2 April 2001. This station continuously monitors di¡use CO2 degassing at Aguas Agrias (785 m a.s.l., Fig. 1), where intense fumarolic and di¡use CO2 degassing occurs. The station is equipped with an on-board microcomputer as well as sensors to measure the soil CO2 gas concentration, wind speed and direction, air temperature and relative humidity (1 m above the ground), and soil water content and temperature (0.3 m depth). Soil CO2 e¥ux is estimated according to the accumulation chamber method [16] by means of a NDIR (non-dispersive infrared) spectrophotometer (Dra«ger Polytron IR transmitter). The reproducibility for the range 10^20 000 g m32 day31 is U 10%. We assumed a random error of U 10% in the emission rates of CO2 , based on variability of the replicate measurements carried out on known CO2 e¥ux rates in the laboratory. Values of CO2 e¥ux (g m32 day31 ) are estimated from the rate of concentration increase in the chamber at the observation site, accounting for changes of atmospheric pressure and temperature to convert volumetric concentrations to mass concentrations. All the recordings are stored on £ash memory within the geochemical station (WEST Systems) and radio-telemetered to Verapaz Mayor’s O⁄ce 5 km away. The observation site is located in a nonvegetated area with highly weathered and compacted soils. Because of hot groundwater discharge and fumarolic steam condensation, soil horizon at 30 cm depth is always damp with a soil water content of over 95% during the observation period.
3. Results and discussion Fumarolic gas samples were collected both from Aguas Agrias and El In¢ernillo at the San Vicente volcano. CO2 is the most abundant component in the dry gas ( s 90 vol%). Aguas Agrias fumaroles displayed a N13 C(CO2 ) = 31.92x and the highest 3 He/4 He ratio, 6.90 U 0.11RA , where RA is the atmospheric 3 He/4 He ratio (1.4U1036 ). El In¢ernillo fumaroles showed N13 C(CO2 ) = 31.98x and a 3 He/4 He ratio of 4.75 U 0.41RA .
EPSL 6456 28-11-02
84
J.M.L. Salazar et al. / Earth and Planetary Science Letters 205 (2002) 81^89
May 8, 2001 earthquake, M=5.1 2500
-1
d -2
gm
CO 2 efflux
2000
Pre-, co-, and post-seismic CO 2 efflux changes
1500 1000 500
Barometric Pressure HPa
0 924
920
916
912
Number of Earthquakes
80 60 40 20 0
Wind speed m/s
6 4 2 0
Apr-01
May-01
May-31
Jul-01
Jul-31
Aug-31
2001 Fig. 2. Time series plot of di¡use CO2 e¥ux, barometric pressure, number of earthquakes from April to August 2001, and wind speed at the San Vicente volcano, El Salvador. Black lines depict 24 h moving averages.
These isotopic compositions suggest a common magmatic origin for the £uids discharged at the San Vicente volcano. A time series of 3667 hourly observations of the di¡use CO2 e¥ux is shown in Fig. 2. Spectral
analysis of this time series yielded a typically inverse correlation with barometric pressure at semi-diurnal and diurnal cycles (12 and 24 h periods, respectively). Spectral coherence between diffuse CO2 e¥ux and wind speed, air temperature
EPSL 6456 28-11-02
J.M.L. Salazar et al. / Earth and Planetary Science Letters 205 (2002) 81^89
85
2500
C{
5000
2000
CO 2 efflux (g m
-2
-1
d )
D
4000
A 1500 3000 1000
2000
B
500
0
1000
(i)
(ii)
(iii)
(iv)
abr-16 may-01 may-16 may-31 jun-16 jul-01Jul-01 jul-16 Apr-01 May-31 May-01
0
Accumulated Released Seismic Energy (ARSE, ergs)
6000
jul-31Jul-31 ago-16 ago-31 Aug-31
2001 Fig. 3. Time series plot of di¡use CO2 e¥ux at Aguas Agrias, San Vicente volcano, and estimated ARSE from April to August 2001. Solid line indicates the CO2 e¥ux time series as 24 h moving average and dots represent the actual hourly measurements.
and air relative humidity is also observed, yielding signi¢cant peaks at 12 and 24 h. These results suggest that short-term £uctuations in the emission rate of carbon dioxide are partially driven by meteorological parameters, as has been similarly described by Rogie et al. [17] at Mammoth Mountain, CA, USA, and Mori et al. [18], at Usu volcano, Japan. Observed low-amplitude temporal variations of the soil temperature (ranging from 91.3 to 94.4‡C) are the result of the bu¡ering e¡ect of low ground permeability with almost constant soil water content. Soil temperature showed a peak of maximum spectral coherence with air temperature at 24 h lag, showing the transfer of solar energy to the soil. These daily variations in solar heating and cooling in£uence atmospheric pressure and short-term £uctuations of carbon dioxide degassing at the observation site. Wind speed showed a median value of 0.3 m/s with a standard deviation of 0.58 m/s (Fig. 2). These low wind speed levels let us ignore the wind speed to explain the following CO2 e¥ux variations on the medium and long-term time
scales. Medium-term £uctuations of the di¡use CO2 degassing from April to August 2001 (Fig. 3) can be divided into four trend windows: (i) quasi-stationary period (V1000 g m32 day31 ) with £uctuations from 637 to 1400 g m32 day31 ; (ii) sustained increase in the mean level (V1500 g m32 day31 , from 644 to 2029 g m32 day31 ); (iii) sharp increasing and decreasing trends (from 239 to 2526 g m32 day31 ); and (iv) quasi-stationary period, with a mean level of V500 g m32 day31 , from 147 to 1272 g m32 day31 . Local seismic activity at the San Vicente volcano is described in terms of the accumulated released seismic energy (ARSE) during the observation period. ARSE was computed summing the log-seismic energy corresponding to each seismic event. Seismic energies reported by Centro de Investigaciones Geote¤cnicas, El Salvador, and calculated using the Richter^Gutenber equation (linear relationship between the logarithm of the seismic energy and the magnitude of an earthquake) were used in this work. Four relevant steps are clearly
EPSL 6456 28-11-02
86
J.M.L. Salazar et al. / Earth and Planetary Science Letters 205 (2002) 81^89
Fig. 4. (A) Plot of di¡use CO2 e¥ux time series showing a precursory signature 8 days (dashed circle) before the onset of a seismic swarm beneath the San Vicente volcano on 8 May 2001. Dashed lines indicate standard deviations from the mean level (1c, s). Arrows show the main steps of the ARSE over the observation period. (B) Plot of di¡use CO2 e¥ux time series showing a delayed response between the changes of di¡use CO2 emission rates and the beginning of the seismic swarm. Arrows show the main steps of the ARSE over the observation period.
EPSL 6456 28-11-02
J.M.L. Salazar et al. / Earth and Planetary Science Letters 205 (2002) 81^89
identi¢ed in the ARSE time series (Steps A, B, C, and D in Fig. 3). Step A relates to the occurrence of a swarm of 18 events in April 2001 centered about 17 km away from the observation site. Step B relates to an intense seismic swarm with a 5.1 magnitude earthquake registered on 8 May 2001, and centered about 20 km away from the observation site. Earthquakes centered within the same epicenter area recorded on 13^15 May 2001 relate to Steps C on the ARSE (which include four minor steps). Step D, with similar characteristics as those described above, occurs at the end of the swarm on 24 May 2001. A signi¢cant relationship between pre-seismic and co-seismic activity and temporal variations in CO2 e¥ux has been observed. Weather conditions were dry during the observation period. Because of the small amount of data at the beginning of the experiment, ARSE Step A is di⁄cult to interpret and will not be discussed. Between the time of occurrence of seismic Steps A and B, a sustained increase of more than 25% ( s 1.0U standard deviation) in the average di¡use CO2 e¥ux 8 days before the 8 May 2001 5.1 magnitude earthquake is observed (Fig. 4A). A detailed analysis of the time series of CO2 e¥ux and ARSE at Step B (Fig. 4B) shows a time shift of 80^90 h between the magnitude 5.1 8 May 2001 earthquake, and a two-fold CO2 e¥ux increase. Oscillations of CO2 e¥ux are up to three times the oscillations during the previous time period. Four short-term £uctuations in the CO2 e¥ux ( 6 12 h) before Steps C in ARSE are also observed (Fig. 4B), with the exception of the second Step C, where the ARSE is ambiguously temporarily related to either an increase or decrease on CO2 e¥ux. It is not easy to distinguish the preseismic from the co- and post-seismic response of CO2 e¥ux at Steps C because dozens of low-magnitude earthquakes shook the area during this period. It is possible that variations of soil CO2 e¥ux at Steps C occurred as a post-seismic response related to the largest seismic event (the M = 5.1 earthquake) and not due to the smaller seismic events of Steps C. An increasing trend of CO2 e¥ux initiated more than 48 h before the ARSE Step D is also observed. The observed changes on the CO2 e¥ux at dif-
87
ferent time scales ranging from days (ARSE, Steps B and D) to hours (ARSE, Steps C) may be considered as precursory earthquake signals. These results suggest that £uid movement and release of CO2 are related to the dynamics of stress accumulation and seismic swarm generation at the San Vicente area. Variations in groundwater pressure produced by changes in crustal stresses can generate £uid con¢nement or injection of deep £uids into the local fault system, a¡ecting CO2 degassing from the water table to the atmosphere. In addition, during time periods (i) and (iv) (Fig. 3) the mean values of CO2 e¥ux were di¡erent. These di¡erences are not physically related to in¢ltration events because it did not rain in El Salvador during this observation period. We suggest two explanations for this behavior: either soil CO2 e¥ux during April 2001 could be considered anomalous with regard to the background levels of period (iv) or the relatively low CO2 e¥ux values observed from June to August 2001 re£ect long-term changes in the hydrodynamic response of the faulting system, caused by the seismic swarm of May 2001. The ¢rst explanation implies that the high seismicity of the area that started in January probably generated background levels of CO2 e¥ux higher than normal and that the levels observed after June are more representative of the normal background levels. Monitoring of CO2 e¥ux and seismicity over a longer period of time could allow a better de¢nition of background levels of CO2 degassing at the San Vicente volcano. It has been widely accepted that £uids play an important role in faulting mechanisms and the triggering of earthquakes [6,19,20]. Independent of the mechanism responsible for its generation, high pore pressure con¢ned within the seimogenic zone will cause pore £uid to di¡use to surrounding areas, leading to fault slippage and seismicity [21]. Seismic tomography studies have recently shown the existence of overpressurized £uids and their e¡ect on enhancing the stress concentration beneath the seismogenic layer, leading to mechanical failure and nucleation of the Kobe earthquake in Japan on 17 January 1995 [22]. Nur and Booker’s previous work [23] suggests that a stress re-distribution after the 6.6 magnitude earthquake
EPSL 6456 28-11-02
88
J.M.L. Salazar et al. / Earth and Planetary Science Letters 205 (2002) 81^89
west of the San Vicente volcano on 13 February 2001 might have occurred, driving £uids through the porous rock, causing changes in pore pressure, and triggering the seismicity recorded on 8 May 2001. Dilatation of the seismogenic zone may have generated an intermittent £ow of CO2 -rich £uids that enhanced gas^water partitioning and fault weakening [24,25]. Carbon and helium isotope signatures reveal a signi¢cant mantle component for the £uids discharged at Aguas Agrias. These ¢ndings suggest that the episodic release of deep magmatic £uids might be an additional source contributing to changes in the local stress ¢eld [26]. Pore pressure changes induced by either changes in stress accumulation within the local faulting system and/or episodic injection of deeply sourced gases (mixing of mantle and metamorphic derived rich-CO2 £uids) might account for our observations on the di¡use CO2 soil e¥ux transient behavior. Further observations are in progress to test our conceptual model of the system and to calibrate the sensitivity of this observation site to seismic and volcanic events. These results show that the application of geochemical methods to monitor volcanic and seismic activity worldwide is promising, and could contribute to the mitigation of seismic and volcanic risks in areas such as the San Vicente volcano.
Acknowledgements We are grateful to the O⁄ce of the Spanish Agency for International Cooperation (AECI) in El Salvador, the Spanish Embassy in El Salvador, The Ministry of Environment and Natural Resources of the Government of El Salvador, and Geote¤rmica Salvadoren‹a (GESAL) for their assistance during the ¢eldwork. We are indebted to the GRP of El Salvador’s Civil National Police for providing security during our stay in El Salvador and John Murray for help with English. Drs. John Rogie and Giovanni Chiodini are also thanked for their helpful reviews of this manuscript. This research was mainly supported by the Spanish AID Agency (Agencia Espan‹ola de Cooperacio¤n Internacional ^ AECI), but addi-
tional ¢nancial aid was provided by the Cabildo Insular de Tenerife, Caja ^ Canarias (Canary Islands, Spain), the European Union, and The University of El Salvador.[BOYLE]
References [1] R.J. Geller, Earthquake prediction: a critical review, Geophys. J. Int. 131 (1997) 425^450. [2] R.J. Geller, D.D. Jackon, Y.Y. Kagan, F. Mulargia, Earthquakes cannot be predicted, Science 275 (1997) 1616^1617. [3] T. Gold, S. Soter, Fluid ascent through the soil lithosphere and its relation to earthquakes, Pure Appl. Geophys. 122 (1984/85) 493^530. [4] R. Sugisaki, T. Ito, K. Nagamine, I. Kawabe, Gas geochemical changes at mineral springs associated with the 1995 southern Hyogo earthquake (M = 7.2), Japan, Earth Planet. Sci. Lett. 139 (1996) 239^249. [5] C.-Y. King, N. Koizumi, Y. Kitagawa, Hydrogeochemical anomalies and the 1995 Kobe earthquake, Science 269 (1995) 38^39. [6] H. Wakita, Geochemical challenge to earthquake prediction, Proc. Natl. Acad. Sci. USA 93 (1996) 3781^3786. [7] P.G. Silver, H. Wakita, A search for earthquake precursors, Science 273 (1996) 77^78. [8] Bull. Global Volcanism Network 26(4) (2001) 9^11. [9] R.E. Torres, C.D. Escobar, D.A. Herna¤ndez, Actividad s|¤smica en San Vicente entre marzo y abril de 1999, Informe Centro Investigaciones Geote¤cnicas, El Salvador (1999). [10] S.G. Rotolo, F. Castorina, Transition from mildy-tholeiitic to calc-alkaline suite: the case of Chintepec volcanic centre, El Salvador, Central America, J. Volcanol. Geotherm. Res. 86 (1998) 117^136. [11] A. Aiuppa, M.L. Carapezza, F. Parello, Fluid geochemistry of the San Vicente geothermal ¢eld (El Salvador), Geothermics 26 (1997) 83^97. [12] T.M. Gerlach, E.J. Graeber, Volatile budget of Kilauea volcano, Nature 313 (1986) 273^277. [13] P.A. Herna¤ndez, K. Notsu, J.M.L. Salazar, T. Mori, G. Natale, H. Okada, G. Virgili, Y. Shimoike, M. Sato, N.M. Pe¤rez, Carbon dioxide degassing by advective £ow from Usu volcano, Japan, Science 292 (2001) 83^ 86. [14] J.P. Toutain, J.C. Baubron, Gas geochemistry and seismotectonics: a review, Tectonophysics 301 (1999) 1^27. [15] W.P. Irwin, I. Barnes, Tectonic relations of carbon dioxide discharges and earthquakes, J. Geophys. Res. 85 (1980) 3115^3121. [16] G. Chiodini, R. Cioni, M. Guidi, B. Raco, L. Marini, Soil CO2 £ux measurements in volcanic and geothermal areas, Appl. Geochem. 13 (1998) 543^552. [17] J.D. Rogie, D.M. Kerrick, M.L. Sorey, G. Chiodini, D.L. Galloway, Dynamics of carbon dioxide emission at Mam-
EPSL 6456 28-11-02
J.M.L. Salazar et al. / Earth and Planetary Science Letters 205 (2002) 81^89
[18]
[19]
[20]
[21] [22]
moth Mountain, California, Earth Planet. Sci. Lett. 188 (2001) 535^541. T. Mori, K. Notsu, P.A. Herna¤ndez, J.M.L. Salazar, N. Pe¤rez, G. Virgili, Y. Shimoike, Hiromu Okada, Continuous monitoring of soil CO2 e¥ux from the summit region of Usu volcano, Japan, J. Volcanol. Soc. Jpn. 47 (2002) 339^345. J. Noorishad, A. Witherspoon, Can injection tests reveal the potential for fault movements? Pure Appl. Geophys. 122 (1984/85) 608^618. A. Spica¤k, J. Hora¤lek, Possible role of £uids in the process of earthquakes swarm generation in the West Bohemia/Vogtland seismoactive region, Tectonophysics 336 (2001) 151^161. X.J. Shi, C.Y. Wang, Instability on a weakening fault, Pure Appl. Geophys. 122 (1984/85) 478^491. D. Zhao, H. Kanamori, H. Negishi, D. Wiens, Tomography of the source area of the 1995 Kobe earthquake:
[23] [24]
[25]
[26]
89
evidence for £uids at the hypocenter?, Science 274 (1996) 1891^1894. A. Nur, J.R. Booker, Aftershocks caused by pore £uid £ow?, Science 175 (1972) 885^887. M. Guerra, G. Etiope, E¡ects of gas-water partitioning, stripping and channelling proceses on radon and helium gas distribution in fault areas, Geochem. J. 33 (1999) 141^ 151. R.H. Sibson, J.M. Moore, A.H. Rankin, Seismic pumping - a hydrothermal £uid transport mechanism, J. Geol. Soc. London 131 (1975) 653^659. N.M. Pe¤rez, J.M.L. Salazar, G. Igarashi, H. Wakita, S. Nakai, Precursory geochemical signatures of recent earthquakes in Galicia, Spain, in: Frontiers in Monitoring Science and Technology for Earthquake Environments, Tono, Japan, Tono Geoscience Center, JNC, 1998, pp. W3^2.
EPSL 6456 28-11-02
