Novel technique to detect seismic signals and its application to map upper-mantle discontinuities beneath Iberia
Resumen Abstract Índice Conclusiones
Anahí Luciana Bonatto
2014-A
En este estudio analizamos las discontinuidades de la zona de transición del manto superior a 410 km y 660 km de profundidad a partir de la detección de ondas sísmicas convertidas de P-a-s debajo de la región Ibero-Magrebí. Para este propósito, usamos eventos telesismicos registrados en 259 estaciones de banda ancha desplegadas mayormente por el proyecto TopoIberia. El análisis detallado de las discontinuidades de la zona de transición proporciona información acerca de la temperatura y composición del manto superior a las profundidades estudiadas. Este estudio añade nuevas restricciones para la mejor comprensión de la compleja y controversial región Ibero-Maghrebí.
Las ondas convertidas en las discontinuidades del manto superior llegan en la coda de la onda P junto con otras señales y usualmente son identificadas en los stacks de funciones receptoras. Aquí, construimos una técnica nueva de procesamiento, que se apoya en las funciones receptoras y que se basa en técnicas de correlación cruzada y de stacking para detectar y extraer señales de manera eficiente a partir de su coherencia, lentitud, tiempo de viaje y polaridad. A fin de añadir consistencia y robustez a las detecciones, nuestros resultados finales se basan en el análisis conjunto de las funciones receptoras y dos funcionales diferentes de correlación cruzada. Esto permite evaluar errores y rellenar gaps en las observaciones cuando alguna de las técnicas falla inherente a las características de la señal y el ruido. Finalmente, la profundidad de las discontinuidades se determina utilizando correcciones de tiempo obtenidas a partir de un modelo de velocidades 3D. Así, presentamos mapas topográficos de las discontinuidades 410-km y 660-km, que muestran variaciones en el espesor de la zona de transición debajo del área de estudio.
El espesor de la zona de transición debajo de Iberia central (240-250 km) está dentro del promedio global; la zona de transición es más ancha debajo del oeste de Marruecos (250-275 km), el Mar de Alborán (280-300 km) y el este de España (260-280); y es predominantemente fina debajo del sur de Portugal (220-240 km), el Golfo de Cádiz (220-250 km) y el área del Estrecho de Gibraltar (214 km). La zona de transición más ancha debajo del oeste de Marruecos y el este de España es mayormente debido a que la discontinuidad 660-km se encuentra a una profundidad mayor que el promedio global, mientras que la topografía de la discontinuidad 410-km es más suave. Aunque, debajo del este de España, se aprecia una leve depresión de la 410. Por otro lado, la profundidad de las discontinuidades está anti-correlacionada debajo del Mar de Alborán. Además, encontramos una correlación espacial entre el vulcanismo anorogénico Neógeno y la topografía de la 410. Todos estos resultados se discuten con el fin de añadir nuevas restricciones a la temperatura y composición de las anomalías de velocidad sísmica observadas en la zona de transición debajo de la controversial región Ibero-Magrebí. El ensanchamiento de la zona de transición del orden de 50 km -respecto al valor de referencia- debajo del Mar de Alborán sugiere que la loza de Alborán aún está lo suficientemente fría como para elevar la transformación de fase alpha-beta y para deprimir la post-spinel. De forma similar, creemos que la loza del Tethys debajo de España -estancada en la base de la zona de transición- aún estaría fría y sería responsable de la depresión de la 660, mientras que un proceso de convección de pequeña escala encima de la 660 -activada por deshidratación de la loza- explicaría la depresión de la 410. Por otro lado, la zona de transición más ancha debajo de Marruecos es probablemente de origen composicional. La explicación preferida es que la depresión de la 660 se debe a la transición granate-a-perovskita sumado a un alto contenido de aluminio en el granate. La zona de transición más angosta debajo del Golfo de Cádiz, el Estrecho de Gibraltar y el sur de Portugal es mayormente debido a una 410 más profunda y pensamos que podría estar causada por un manto superior de elevada temperatura, que también ha sido inferido en imágenes tomográficas recientemente publicadas.
Adicionalmente, determinamos el espesor de las discontinuidades 410-km y 660-km e investigamos su variación espacial. Este análisis, muestra que ambas discontinuidades presentan variaciones espaciales en su espesor. En particular, la 660 es más ancha debajo del Mar de Alborán y el sur de España. Interpretamos la variación espacial en el espesor de la 410 como causada por variaciones en la concentración de agua dentro de la zona de transición debajo del área de estudio. Creemos que la 660 más ancha, de aproximadamente 30 km, debajo del Mar de Alborán y del sur de España es causada por la combinación de gradientes de velocidades debido a las transformaciones de fase post-spinel e ilmenita-a-perovskita.
In this study, we analyze the upper-mantle transition zone discontinuities at a depth of 410 km and 660 km as seen from seismic P-to-s wave conversions beneath the Ibero-Maghrebian region. For this purpose, we use teleseismic events recorded at 259 broadband seismic stations deployed mainly by the TopoIberia project. The detailed analysis of the transition-zone discontinuities provides information on the temperature and composition of the upper mantle at the investigated depths. This study adds new constraints, which would help to improve the understanding of the complex and controversial Ibero-Maghrebian region.
The converted waves from the upper-mantle discontinuities arrive in the P-wave coda together with other signals and are usually identified on stacked receiver functions. Here, a new processing approach is built, which is leaned on receiver functions and which is based on cross-correlation and stacking techniques, to efficiently detect and extract signals by means of their coherence, slowness, travel time and polarity. In order to add consistency and robustness to the detections, our final results are based on a joint analysis of the receiver functions and two different cross-correlation functionals. This permits to assess errors and to bridge observation gaps due to detection failure of any of the techniques inherent to signal and noise characteristics. Finally, discontinuity depths are determined using time corrections obtained from a 3D velocity model. We present topography maps for the 410-km and 660-km discontinuities, which show variations in the transition zone thickness beneath the study area.
The transition zone thickness is about global average beneath central Iberia (240-250 km); it is thicker beneath west Morocco (250-275 km), the Alboran Sea (280-300 km) and east Spain (260-280); and it is predominantly thinner beneath south Portugal (220-240 km), the Strait of Gibraltar area (214 km) and the Gulf of Cadiz (220-250 km). The thicker transition zone beneath west Morocco and east Spain is mainly due to a deeper 660-km discontinuity, while the topography of the 410-km discontinuity is smaller. Although, beneath east Spain, the 410 is slightly depressed. On the other hand, the depth of the discontinuities are anti-correlated beneath the Alboran Sea. Additionally, we find a spatial correlation between the Neogene anorogenic volcanism and the topography of the 410-km discontinuity. These results are discussed to add new constraints on temperature and composition to seismic velocity anomalies observed in the transition zone beneath the controversial Ibero-Magrhrebian region. The transition zone thickening of about 50 km -from the reference value- beneath the Alboran Sea suggests that the Betic-Alboran slab is still sufficiently cold to elevate the alpha-beta mineral phase transition and to depress the post-spinel one. Similarly, the cold Tethys slab -stagnant at the base of the transition zone- beneath east Spain is thought to be responsible for the 660 depression, while small-scale convection above the 660 -triggered by slab dehydration- may explain the 410 depression. On the other hand, the thicker transition zone beneath Morocco is probably of compositional origin. Our preferred explanation is that the 660 depression is due to the garnet-to-perovskite transition and a high aluminum content within garnet. The thinner transition zone beneath the Gulf of Cadiz, the Strait of Gibraltar and the south of Portugal is mainly due to a depressed 410 and is thought to be caused by high upper-mantle temperature, which is also inferred by recently published tomographic images.
Furthermore, we determine the widths of the 410-km and 660-km discontinuities and we investigate their spatial variations. This analysis has revealed that both discontinuities present spatial thickness variations. In particular, the 660 is thicker beneath the Alboran Sea and south Spain. We interpret the spatial variation of the 410 width as caused by variations in the water concentration in the transition zone beneath the study area. The thicker 660, of about 30 km, beneath the Alboran Sea and south Spain is thought to be caused by combined velocity gradients due to post-spinel and ilmenite-to-perovskite phase transitions.
1 General introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1 Motivation and organization of the thesis . . . . . . . . . . . . . . . 4
1.2 Upper mantle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.1 Upper mantle composition . . . . . . . . . . . . . . . . . . . . . . 10
1.2.2 Olivine-related TZ discontinuities . . . . . . . . . . . . . . . . . 12
1.2.3 Using the 410 and 660 depths to infer changes in TZ temperatures . . 13
1.3 Research methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.1 Seismic phases . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.3.2 Spatial resolution . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.3.3 Detection of P -to-s converted phases in the seismic records . . . . 17
1.4 The western Mediterranean and the Ibero-Maghrebian region . . . . . . . 19
1.4.1 Deep earthquakes beneath Granada . . . . . . . . . . . . . . . . . . 21
1.4.2 Tomographic images of the upper mantle . . . . . . . . . . . . . . . 22
1.4.3 A controversial geodynamic scenario in the Alboran Sea area . . . . 24
1.4.4 Anorogenic magmatism . . . . . . . . . . . . . . . . . . . . . . . . 26
1.4.5 Seismic discontinuity studies . . . . . . . . . . . . . . . . . . . 28
1.5 TopoIberia data set . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2 Methodology: detection of P -coda phases . . . . . . . . . . . . . . . . . 32
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.1 Theoretical background . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.2 Detection of P-coda phases using cross-correlation . . . . . . . . . 36
2.3 Synthetic analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.3.1 Generating synthetic data . . . . . . . . . . . . . . . . . . . . . 46
2.3.2 Pilot length . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.3.3 Noise influence . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.3.4 Robustness analysis . . . . . . . . . . . . . . . . . . . . . . . . 53
2.4 Real data examples . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.4.1 Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.4.2 Detection of P -to-s conversions at individual stations . . . . . . 57
2.5 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . . 64
3 Data set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.2 Data selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.3 Processing and final data set . . . . . . . . . . . . . . . . . . . . . 72
3.4 Building of correlograms and receiver functions . . . . . . . . . . . . 75
3.5 Data consistency . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.5.1 Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.5.2 Pilot length . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.6 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . 79
4 Transition zone discontinuities beneath Iberia and Morocco . . . . . . . . 84
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.2 Data and method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.2.1 Stacking of correlograms and receiver functions . . . . . . . . . . 87
4.2.2 Robustness analysis and quality criteria . . . . . . . . . . . . . . 88
4.2.3 Integrated detections and depth conversions . . . . . . . . . . . . 92
4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.3.1 410 and 660 discontinuities . . . . . . . . . . . . . . . . . . . . 97
4.3.2 Time corrections and the 410 and 660 absolute depths . . . . . . . 105
4.3.3 TZ thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.3.4 Additional features in the receiver functions . . . . . . . . . . . 110
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.4.1 Relation with previous works . . . . . . . . . . . . . . . . . . . 115
4.4.2 Interpretation of results . . . . . . . . . . . . . . . . . . . . . 116
4.5 Summary and conclusions . . . . . . . . . . . . . . . . . . . . . . . 130
5 Discontinuity characterization . . . . . . . . . . . . . . . . . . . . . 134
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.2 Relative amplitudes . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.2.1 Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.2.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
5.3 Width of the 410 and 660 discontinuities . . . . . . . . . . . . . . . 141
5.3.1 Methodology and processing . . . . . . . . . . . . . . . . . . . . 142
5.3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
5.3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Appendix A Receiver functions . . . . . . . . . . . . . . . . . . . . . . . 158
A.1 Transmission path impulse response . . . . . . . . . . . . . . . . . . 160
A.2 Water-level deconvolution receiver functions . . . . . . . . . . . . . 162
Appendix B The presence of other transforming and non-transforming
phases and their geophysical implications . . . . . . . . . . . . . . . . . 164
B.1 Garnet-related discontinuities near 660 km depth . . . . . . . . . . . 166
B.2 410 and 660 complexities . . . . . . . . . . . . . . . . . . . . . . . 167
B.3 Influence in the TZ thickness . . . . . . . . . . . . . . . . . . . . 169
B.4 510-km discontinuity . . . . . . . . . . . . . . . . . . . . . . . . . 169
Appendix C Supplementary figures for Chapter 2 . . . . . . . . . . . . . . 172
Appendix D TopoIberia stations . . . . . . . . . . . . . . . . . . . . . . 190
Appendix E Supplementary figures for Chapter 4 . . . . . . . . . . . . . . 200
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
CONCLUSIONS:
CHAPTER 2: Methodology.
We have presented a new processing approach based on cross-correlation and stacking techniques to detect weak amplitude phases that arrive in the P phase coda. As cross-correlation functional, we have used the CCGN, which is the classical approach to measure waveform similarity, and the PCC, which is based on instantaneous phase coherence and which has not been widely explored before. We have proposed to use these cross-correlations (PCC and CCGN) together with receiver functions (RFs) to stabilize the detections against errors and to bridge observation gaps. Without noise or with low to reasonable noise level, the results of the three techniques are similar but inherent to their different strategies the signals are detected differently. Similar results obtained through different methods add robustness and confidence to the detections and interpretations. Varying results or non-detections are expected for more difficult data, depending on the signal and noise characteristics. Therefore, the variations of the final results and the amount of approaches which lead to an independent detection can be used as a quality indicator. Our procedure is a step forward to stabilize against detection problems and to identify more ambiguous detections. Another advantage is that one automatically bridges observation gaps by one or the other method. Each difficult detection is a special case due to the non-stationarity of signals and noise and none of the methods will lead to satisfying detections for all cases.
PCC is amplitude unbiased and more sensitive to waveform coherence than CCGN and RF. Coherent signals are therefore accurately detected even in the vicinity of other larger amplitude signals which may bias the detections with CCGN. Our synthetic data tests show that PCC may even provide more stable results than CCGN and RF at high noise levels. Of course, PCC fails when signals cannot be detected by their coherence. In these cases, CCGN is usually the better approach. The water-level deconvolution used to compute the RFs has the inconvenience that the choice of the water-level parameter is a subjective task. However, we have used a fixed water-level parameter of 0.1 and a quality control over the RFs amplitude, which simplify the automatic computation of RFs. In practice, we see that the detection of P410s and P660s in individual records is a more difficult task, which can be deduced from the individual traces in the time-back azimuth sections. We attribute this complexity to the loss of waveform coherence with the reference phase, which is particularly disadvantageous to the PCC. Nevertheless, the P410s and P660s phases are well detected with the three techniques in the time-slowness domain. We observed that the 95 % confidence interval of the estimated time values with each technique are overlapped and that they follow the same trend, regardless of the technique. All this suggests that the measured time values with CCGN, PCC and RF are consistent and that we can rely on the estimated time values from a single technique.
CHAPTER 3: Data set.
The number of records that remain after the quality controls is significantly smaller than the number of events in the original data set. This is not unexpected because the useful data always corresponds to a small percentage of the available data. The increasing number of stations in local and regional networks turns to be a great advantage because it permits to exploit the great potential of the useful records. The 259 stations that we use provide a vast volume of data which leads to an unprecedented data coverage for the study of the 410 and 660 discontinuities and the TZT beneath the Iberian Peninsula and Morocco.
The analysis with the three different frequency bands suggests that the lowest frequency band (0.02-0.12 Hz) is the more adequate to perform the processing of the entire data set since two of the three techniques show a higher number of signals in the lower frequency band. However, the number of detections near P410s and P660s for the three frequency bands and techniques is considerably larger than for other time intervals. This enables us to conclude that any of the three frequency bands proves to be a good choice. This result permits us to base our choice of frequency band on other important aspects, such as time-slowness resolution and spatial resolution.
Additionally, the number of peaks near P410s and P660s for PCC and CCGN does not show strong dependence on the pilot length choice. However, we have seen that for the RFs the smallest differences between the number of detections near P410s and P660s is obtained with the largest pilot. Thus, we consider that the 100 s pilot is the more adequate pilot length to perform the data processing. In general, for both analyses (frequency and pilot length) we have seen that the number of signals near P660s is always larger than near the P410s.
CHAPTER 4: Transition zone discontinuities beneath Iberia and Morocco.
We have carried out a semiautomatic search for converted phases P410s and P660s beneath the Iberian Peninsula and north Morocco using teleseismic data from 259 stations, most of them belonging to the IberArray of the TopoIberia project. We have used three independent approaches (PCC, CCGN and RF) to estimate the relative travel time of converted phases and a bootstrap algorithm to estimate the time uncertainties. The region has been divided in small areas of common piercing points to perform the stack of correlograms and receiver functions. We have considered only converted phases with stacked amplitude larger than twice the mean amplitude in the time interval 30-80 s, and with bootstrap time-standard deviation smaller than 1.5 s. Besides, we have performed a thorough visual inspection to avoid spurious detections. Based on a body wave tomography velocity model for the study area, we have applied time corrections to the travel times of the Pds (d=410, 660) phases. These corrected travel times were then converted to depth obtaining the topography maps for the 410 and 660 discontinuities and the TZT map beneath the Iberian Peninsula and its surroundings.
Our main conclusions are the following:
1- Our results are in good agreement with previous studies in the same area (van der Meijde et al., 2005; Dündar et al., 2011). However, the large data volume provided by the IberArray of the TopoIberia project has permitted us to considerably increase the resolution and to resolve new TZT and topography of discontinuities.
2- Alboran Sea area:
– We have found consistency between the thickened TZ beneath the Alboran Sea and the position of a high-velocity anomaly in the tomographic images which have been related to the Betic-Alboran slab. Our results suggest that the slab is still cold enough to induce downward deflection of the post-spinel transformation and uplift of the olivine-to-wadsleyite phase transition.
– The anti-correlated depth of the 410 and 660 beneath the Alboran Sea have provided indirect and independent (from tomography) evidences confirming the steep nature of the Betic-Alboran slab.
– The nest of deep earthquakes beneath Granada is in an area of high temperature gradient inside the TZ (of about 540 K over a distance of less than 250 km). This may further help to constrain the origin of the deep earthquakes beneath Granada.
3- The depressed 410 beneath the Balearic Sea, the presence of strong Pws phases before P410s and the depressed 660 in the Pyrenees and Balearic Sea are consistent with the presence of the Alpine-Tethys remnant slab. The cold stagnant slab can explain the 660 depression; small-scale convection above the 660, triggered by slab dehydration can explain the depressed 410; and the presence of a water concentration increase in the TZ due to dehydration of the stagnant slab can explain the strong reversed polarity signal before P410s. Nevertheless, other models proposed to explain the anorogenic magmatism in the Mediterranean are also consistent with our observations.
4- We have proposed a compositional origin for the thickened TZ beneath the western Moroccan region. The post-garnet transition together with an Al-rich mantle can explain the downward deflection of the 660 beneath Morocco and the thicker TZ in this area.
5- The decreased visibility of the P410s phase beneath the Rif and beneath the western end of the Alboran Sea is possibly caused by structural complexities and/or the thermodynamic properties of the olivine-to-wadsleyite phase change in colder-than-normal mantle.
6- For the Gulf of Cadiz and the southwest coast of Portugal we have proposed a thermally thinned TZ, which is consistent with the tomographic images from Monna et al. (2013). Besides, a warm environment and the presence of Al within garnet would also explain the small length-scale depth changes of the 660. We attribute the deeper 660 to the garnet-related phase transition and the shallower 660 to the postspinel transition.
7- The presence of melt atop the 410 may explain reversed polarity arrivals from above this discontinuity. Plume material beneath the Gulf of Cadiz and subducted oceanic-lithosphere beneath the Alboran Sea and northeast Spain could provide the higher water content to explain the stronger reversed polarity signal in these areas.
8- We have found that the spatial correlation between the active anorogenic magmatism and the 410 depression beneath Morocco and the northeast coast of Spain can be explained with both groups of models proposed for the origin of this magmatic activity. The estimated 410 depression in this area is consistent with a plume-like model with the steam of the plume away from the Mediterranean and affecting only the upper portion of the TZ or with a Mg-enriched mantle due to decompression melting in the overlaying mantle, which shifts the 410 phase transition to higher pressures.
CHAPTER 5: Discontinuity characterization.
The P410s amplitudes and the 410 widths are consistent with spatial variations in the water concentrations in the TZ beneath Iberia. In particular, the variation of the 410 thickness (7-28 km) agrees with 200-500 ppm of water in wadsleyite. On the other hand, to explain the anomalously high relative amplitude values of the P410s phase (P410s/P) we need to assume a negative velocity jump atop the 410, which increases the total velocity jump across the 410 km depth discontinuity. The negative velocity jump immediately above 410 km depth is consistent with the reversed-polarity signal that we observed before P410s, which is related with partial melt atop the 410 caused by a water concentration increase in the TZ.
The spatial variation of the 660 thickness might reflect the presence of the garnet-related phase transition (garnet-to-perovskite or ilmenite-to-perovskite) together with the post-spinel. This could lead to complex velocity structures which might be responsible for the broader 660 in the cold Alboran Slab. The estimated relative amplitude values of the P660s phase agree with the velocity jump in AK135 at 660 km depth.