Supplementary Figures

Transcription

1 Supplementary Figures Dyrfjöll (Iceland). Complete set including all dykes 5 Log normal Chi -squared Dyrfjöll (Iceland). All mafic dykes 5 Lognormal Chi-squared Dyrfjöll (Iceland). Felsic dykes Lognormal Chi-squared Dyrfjöll (Iceland). Cone sheets, mafic Log normal Chi -squared Dyrfjöll (Iceland). Regional dykes, mafic Lognormal Chi-squared Geitafell (Iceland). Complete set including all dykes 5 Log normal Chi -squared

2 Geitafell (Iceland). All mafic dykes 5 Lognormal Chi-squared Geitafell (Iceland). Cone sheets, mafic Log normal Chi -squared Geitafell (Iceland). Regional dykes, mafic 5 Lognormal Chi-squared La Palma (Canary Island). Complete set including all dykes 5 Log normal Chi -squared La Palma (Canary Islands). Basal complex, all dykes Lognormal Chi-squared La Palma (Canary Islands). Basal complex, mafic dykes only Lognormal Chi-squared La Palma (Canary Islands). Basal complex, felsic dykes only. Log normal Chi -squared

3 5 La Palma (Canary Islands). Taburiente volcano, all dykes Log normal Chi -squared La Palma (Canary Islands). Taburiente volcano, mafic dykes only Lognormal Chi-squared 5 La Palma (Canary Islands). Taburiente volcano, intermediate composition Log normal Chi -squared Tenerife (Canary Islands). Complete set including all dykes Lognormal Chi-squared SUPPLEMENTARY FIGURE : Q-Q plots for all datasets. Ordinate gives the observed dyke thicknesses, abscissa gives the corresponding dyke thickness values calculated from the given distribution. A perfect fit of the statistical distribution to the data would be indicated if all data plot on the 5 line shown. The box in each plot contains 8 % of the data (excluding the thickest dykes).

4 a) W 9 W W 66 N 65 N 6 N N 5 km b) 9 N Rift zone (>.8 Ma) Volcanic system Plio-Pleistocene (.8-. Ma) Central volcano, clear/unclear Tertiary (>. Ma) Glacier 8 W 6 W W 5 km Lanzarote La Palma Tenerife Gran Canaria 8 N La Gomera El Hierro Fuerteventura N SUPPLEMENTARY FIGURE : Location of field areas where dyke thicknesses were measured. a) Geological map of Iceland with the Tertiary central volcanoes Dyrfjöll () and Geitafell () marked. b) Map of the Canary Islands with the location of dyke thickness measurements on La Palma () and the Anaga shield volcano on Tenerife () marked.

5 A) n = 86 max. density = 87.8 min. density =. contour intervals = 5 5'W N B) W N Hoffellsjökull n = 56 max. density = min. density =. contour intervals = 5 5 km 6 5'N C) 7 57 W N Dyrfjöll 65 'N 5 km Taburiente Basal complex n = 98 max. density = 9 min. density =. contour intervals = 5 Taburiente volcano n = 75 max. density =. min. density =.8 contour intervals = 5 8 N 5 km D) W N Las Casas de Afur Anaga massif Igueste n = 55 max. density = 5 min. density =.5 contour intervals = 5 5 km San Andrés 8 N SUPPLEMENTARY FIGURE : Location of profiles in the four field areas represented by red lines. A) Dyrfjöll and B) Geitafell, Iceland. For further details see Burchardt et al. C) Taburiente and Basal Complex (black lines), La Palma, and D) Anaga massif Tenerife, Canary Islands. Insets show location of study area. Lower hemisphere equal-area projection of poles to orientation planes with density contours marked (insets). Contours range from zero (white intervals) to maximum (black intervals). Stereographic projections in A) and B) show regional dykes in red and cone sheets in blue symbols following the subdivision of Burchardt et al. This subdivision appears robust as only a negligible number of dykes classified as regional dykes do not follow the regional strike trend. In the stereographic projections in C) and D), no further subdivision into dyke types is shown as sub-groups are not obvious in the field and from petrology/geochemistry. Hence, all dykes are indicated by blue symbols. Base maps are from ASTER GDEM V data (ASTER GDEM is a product of METI and NASA). 5

6 Supplementary Tables Supplementary Table : Statistical tests for the complete Dyrfjöll (Iceland) dataset including all lithologies and types of dykes. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). The goodness of fit is calculated for the Kolmogorov-Smirnov criterion as GoF(KS) = max c i c(x i ), and for the residual sum of squares as GoF(RSS) = (c i c(x i )) /n, where the c i are the values of the cumulative distribution function (CDF) given by the data, the c(x i ) are the corresponding CDF values of the tested distribution evaluated at x i, and n is the number of data points. n = 87 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par...8. (.7).7 (.) λ =.87 k = () () λ =.89 Lognormal.6.75 () () σ =.85 µ = () () σ =.59 Chi-squared () () k = () () α =. x min =. 6

7 Supplementary Table : Statistical tests for the Dyrfjöll (Iceland) dataset including all dyke types but only mafic lithologies. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par (.77).79 (.7) λ =.876 k = () () λ =.8 Lognormal () () σ =.8 µ = () () σ =.596 Chi-squared.85.5 () () k = () () α =.9 x min =. 7

8 Supplementary Table : Statistical tests for the Dyrfjöll (Iceland) dataset including all dyke types but only felsic lithologies. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 5 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par (.787).6 (.787) λ =.85 k = (.5).99 (.8) λ =.9 Lognormal (.).5 (.57) σ =.99 µ = (.7). () σ =.58 Chi-squared (.). (.) k = () () α =.7 x min =.9 8

9 Supplementary Table : Statistical tests for the Dyrfjöll (Iceland) dataset including only cone sheets of mafic lithology. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 98 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par (.).9 (.9) λ =.85 k =..7.7 () () λ =.57 Lognormal (.) () σ =.86 µ = () () σ =.55 Chi-squared () () k = () () α =.56 x min =. 9

10 Supplementary Table 5: Statistical tests for the Dyrfjöll (Iceland) dataset including only regional dykes of mafic lithology. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par (.75).77 (.) λ =.5 k = (.) () λ =.95 Lognormal (.) () σ =.866 µ = () () σ =.69 Chi-squared () () k = () () α =. x min =.

11 Supplementary Table 6: Statistical tests for the complete Geitafell (Iceland) dataset including all lithologies and dyke types. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 56 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par (.97).7 (.7) λ =.59 k = (.87).5 (.5) λ =.8 Lognormal.8. (.7) () σ =.8 µ = () () σ =. Chi-squared.68.5 () () k = () () α =. x min =.

12 Supplementary Table 7: Statistical tests for the Geitafell (Iceland) dataset including only mafic dykes. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 5 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par (.97).59 (.) λ =.57 k = (.6).67 (.87) λ =.875 Lognormal.5.. (.). (.) σ =.86 µ = () () σ =. Chi-squared () () k = () () α =.5 x min =.

13 Supplementary Table 8: Statistical tests for the Geitafell (Iceland) dataset including only mafic cone sheets. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par....6 (.7).96 (.767) λ =.87 k = (.).95 (.99) λ =.8 Lognormal.8.8 (.) (.) σ =.57 µ = () () σ =.6 Chi-squared () () k =.89.. () () α =. x min =.

14 Supplementary Table 9: Statistical tests for the Geitafell (Iceland) dataset including only regional dykes of mafic lithology. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 9 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par..5.. (.).5 (.) λ =.6 k = (.7).68 (.5) λ =.6 Lognormal (.6).96 (.6) σ =.5 µ = () () σ =.6 Chi-squared.7.75 () () k = () () α =.8 x min =.

15 Supplementary Table : Statistical tests for the complete La Palma (Canary Islands) dataset including all dyke types and lithologies. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the declustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 9 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par..97. () () λ =.99 k = () () λ =.5 Lognormal.65.9 () () σ =.866 µ = () () σ =.65 Chi-squared.88. () () k = () () α =. x min =. 5

16 Supplementary Table : Statistical tests for the La Palma (Canary Islands) dataset including all dyke types and lithologies from the Basal Complex, only. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 98 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par..9. () () λ =.68 k = () () λ =.56 Lognormal.7.9 () () σ =.85 µ = () () σ =.7 Chi-squared () () k = () () α =. x min =. 6

17 Supplementary Table : Statistical tests for the La Palma (Iceland) dataset including only mafic dykes from the Basal Complex. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the declustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 8 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par..85. (.) () λ =.6 k = () () λ =.589 Lognormal.5. () () σ =.86 µ = () () σ =. Chi-squared..6 () () k = () () α =. x min =. 7

18 Supplementary Table : Statistical tests for the La Palma (Canary Islands) dataset including only felsic dykes from the Basal Complex. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 77 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par (.87).567 (.6) λ =.59 k = () () λ =.7 Lognormal (.5). (.) σ =.796 µ = (.). (.7) σ =. Chi-squared.58.9 () () k = () () α =.555 x min =.8 8

19 Supplementary Table : Statistical tests for the La Palma (Canary Islands) dataset including all dykes from Taburiente volcano. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the declustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 75 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par (.). (.9) λ =. k =..9.9 () () λ =.856 Lognormal (.7).59 (.9) σ =.95 µ = () () σ =.7 Chi-squared () () k = () () α =.7 x min =.5 9

20 Supplementary Table 5: Statistical tests for the La Palma (Canary Islands) dataset including only mafic dykes from Taburiente volcano. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par (.5).9 (.) λ =.86 k = () () λ =.9 Lognormal (.).55 (.7) σ =.88 µ = () () σ =.5 Chi-squared.7.68 () () k = () () α =.98 x min =.5

21 Supplementary Table 6: Statistical tests for the La Palma (Canary Islands) dataset including only dykes of intermediate composition from Taburiente volcano. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the de-clustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 55 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par (.97).7 (.98) λ =.7 k = (.7) () λ =.55 Lognormal (.7).5 (.7) σ =.7 µ = (.8).9 (.) σ =.7 Chi-squared (). () k = () () α =.5 x min =.

22 Supplementary Table 7: Statistical tests for the complete Tenerife (Canary Islands) dataset including all dyke types and lithologies. Goodness of fit (GoF), p-values, and parameter values for the best fitting distributions are reported. The p-values are calculated for the unmodified datasets as well as for the declustered datasets (in parentheses). Goodness of fit is calculated as in Supplementary Table. n = 55 GoF (KS) GoF (RSS) P (KS) P (RSS) Par. Par (.7). (.) λ =.7 k = (.7).9 (.) λ =.57 Lognormal.7.5 () () σ =.87 µ = () () σ =. Chi-squared (.). () k = () () α =. x min =.

23 Supplementary Note : Detailed derivation of distribution in dyke thicknesses The derivation of distributed failure stresses from power law distributed flaws is presented elsewhere. The necessary condition that the probability of failure of any given flaw is small is clearly satisfied here. Fracturing by overpressure is mechanically equivalent to fracturing by tensile stress because the stress concentrations at the stress tip are equal,. The overpressures in the magma chamber when new dykes are initiated ( P ) thus follow the same distribution as the failure stresses. Hoop stresses in the magma chamber wall due to chamber inflation are linearly related to the overpressure 5. Hoop stresses are tensile and hence reduce the overpressure required for dyke nucleation, but due to the linear relation between overpressure and hoop stress, this mechanism affects only the λ of the distribution, but not the k. The excess magma volume in the chamber is proportional to the initial overpressure. Excess magma may be accommodated by chamber inflation, surface updoming, or compression of the magma 6. Since the excess volume is only a small perturbation to the chamber volume, chamber inflation increases linearly with pressure, regardless of the chamber geometry. Updoming and magma compression are also to first order linear functions of overpressure, so that V chamber P () where V chamber is the excess volume of magma in the magma chamber before dyke initiation. An important condition for the linearity of magma compression is clearly that the compressibility of the magma remains constant over the time period of by the dyke swarm. This condition requires approximately constant volatile pressure and hence more or less constant magma temperature and composition. The excess volume then follows a distribution with the same k as that of the failure stresses. The final overpressure is low when dyke propagation ceases (or at least the overpressures are similar after completion of different dyking events). This assumption is most likely justified since large dykes require less overpressure for propagation than small ones, and the final dyke sizes are relatively similar when compared with the initial flaws. As a result, the dyke volume is the same as the initial excess volume of magma in the chamber because all excess magma migrates into the dyke, and because the magma in the dyke at the end of dyke propagation is not significantly compressed. With V chamber V dyke, the dyke volumes are distributed with the samek as for failure stresses. The individual dyke volume is proportional to a power of the dyke thicknessh, that is,v dyke h m. Assuming a dyke that is half of an ellipsoid (with the thick end located at the contact with the magma chamber) with semi-major axes c and semiminor axis h, the volume of the dyke is V dyke = πc h. We note that dykes in nature are generally not perfect ellipsoids 7 and other shapes such as a disc with constant thickness might better approximate the shape. The important aspect for our model is that all dykes in a given swarm have the same final shape. In equilibrium, the dyke s aspect ratio is determined by the overpressure,h/c P. The condition for dyke propagation, which equally describes the condition when the dyke ceases to propagate, provides an additional relation between dyke thickness, length, and overpressure, so that the dyke volume can be written in terms of thickness alone. Dyke propagation depends on the mechanics of the dyke tip, for which different models exist. The first dyke tip model is based on linear elastic fracture mechanics 8 (LEFM): The stress con-

24 centration factor K is a function of the tensile stress (or overpressure) and dyke length, K = P πc (Lawn and Wilshaw 8, eq..7) () where P is the overpressure during or upon termination of dyke propagation. Propagation occurs when the total elastic energy released by fracture elongation is greater than or equal to the energy consumed in creating new surfaces K E = γ (Lawn and Wilshaw 8, eq..) () where E is Young s modulus of the rock, and γ is the free surface energy. Since E and γ are constants, so is K, and the propagation criterion yields P /c, indicating that long dykes require less overpressure for propagation than short ones. Combining equation () with the relations for dyke volume and dyke aspect ratio, the dyke volume is found to be V dyke h m with m = 5. As an aside, the model prediction can be rewritten as dyke thickness vs. dyke length: h c.5. This result is in very good agreement with measured thickness to length ratios for which a best fit yields 9 h c.8. The second type of dyke tip model comprises the Barenblatt model of an elastic fracture tip with a given cut-off stress 8, and the Dugdale model of a plastic fracture tip 8. For these two models, an analogous argument can be made as for the LEFM model. The free surface energy γ is again constant, thus leading tov dyke h 5. A third type of dyke tip model is applicable to hydrofracture under large confining stress. For such dykes, the dynamics are believed to differ significantly from the LEFM, Barenblatt, and Dugdale models introduced above. The damage zone is no longer confined to thek-dominant region. Fracture toughness is then not a material property but scales with crack length. As a result, the overpressure required for propagation is independent of crack length, and hence V dyke h. CDF, The distribution of h m is another distribution of h. This fact is obvious for the ( C(h m ) = exp h mk) ( = exp h k ) () but for the PDF attention must be paid to the fact that the resulting probability density must be referenced to h instead of h m, that is, to the probability per unit length: P(h m ) = (hm ) ( h P m(h m ) = mh m Ah mk m exp h mk) ( = Bh k exp h k ) (5) where P m is the probability density referenced to h m, which results from simply substituting h m for h in the original PDF. Given that the dyke volumes are distributed with a shape parameter k, it follows that the dyke thicknesses follow a distribution with a parameter k = mk, where m = 5 for the LEFM, Barenblatt, and Dugdale models, and m = for dyke formation under high overburden.

25 Supplementary Note : Detailed geological setting Increased heat flow and magmatic activity caused by the Iceland mantle plume, which is superimposed on the Mid-Atlantic ridge, has resulted in the formation of the Iceland plateau, where mid-ocean ridge processes occur above sea level within volcanic zones. Within the volcanic zones, crustal spreading occurs along en-echelon aligned volcanic systems and is accommodated by normal faulting in the uppermost few hundred metres of the crust and by magmatic sheet intrusions (dykes) at depth. Regional dykes feed fissure eruptions that give rise to flood basalts, that have built the upper crust in Iceland. Central volcanic complexes within the volcanic zone are fed from shallow magma chambers, which is why the occurrence of felsic rocks is generally restricted to the central volcanic complexes. The extinct parts of the rift zone of Iceland are deeply eroded and exposed e.g. in eastern and western Iceland. There, glacial valleys and fjords expose sections through the uppermost two kilometres of the crust and offer insight into the subsurface structure of the rift zone. Vertical regional dykes that cut and acted as feeders of the flood basalts are arranged in swarms, in which the density of sheet intrusions increases with depth. Generally, the fossil volcanic systems comprise one central volcanic complex and regional dykes. The deeply-eroded glacial valleys grant access to the interior of central volcanic complexes, which usually consists of a pluton that represents the remnant of the shallow magma chamber surrounded by swarms of inclined sheets. We measured the thickness of dykes in two of the extinct and deeply-eroded central volcanoes in Eastern Iceland (Supplementary Fig. and ), namely Geitafell in Southeast Iceland 5 and Dyrfjöll in Northeast Iceland,6. The Canary archipelago, located in the central North Atlantic, ca. to 5 km off NW-Africa, consists of seven islands that were volcanically constructed during the last to Ma 7. Magmatic activity is related to an underlying mantle plume with rift zones and central volcanic complexes as the surface manifestations of this activity 8. The evolution of the Canary Islands follows a typical succession of stages that include the seamount stage with subaquaceous volcanism, the shield stage during which a shield volcano was constructed above sea-level, the post-shield stage with decreased magma supplyrates, and the post-erosional stage with renewed volcanic activity 9. The thickness of dykes was studied on the islands of Tenerife and La Palma (Supplementary Fig. and ). On La Palma, we measured the thickness of magmatic sheets within the Basal Complex that represents a deeply eroded section through the seamount stage of the island. Additional dyke thickness measurements were recorded along the crest and steep inner slopes of the Caldera de Taburiente depression that exposes the interior of the Taburiente volcanic edifice. The Taburiente volcano comprises several successive volcanic shields 9, that consist of a stack of interlayered lava flows and pyroclastic rocks. Measurements of dyke thicknesses on the island of Tenerife were carried out along road cuts that expose the interior of the extinct Anaga shield volcano in the northeast of the island. The Anaga volcano developed from an initially ridge-like volcanic edifice to a triple-armed rift system some. Ma ago and is well-exposed due to deeply-eroded canyons and major collapse embayments. 5

26 Supplementary References [] Burchardt, S., Tanner, D.C., Troll, V.R., Krumbholz, M. & Gustafsson, L.E. Three-dimensional geometry of concentric intrusive sheet swarms in the Geitafell and the Dyrfjöll Volcanoes, Eastern Iceland. Geochem. Geophys. Geosys., QAB9, doi:.9/gc57 (). [] Freudenthal, A.M. Statistical approach to brittle failure, Fracture, Vol., ed. H. Liebowitz, Academic Press, (968). [] Rubin, A. M. Propagation of magma-filled cracks. Annu. Rev. Earth Planet. Sci., 87-6 (995). [] Pollard, D.D. & Segall, P. Theoretical displacements and stresses near fractures in rock: with applications to faults, joints, veins, dikes, and solution surfaces. In: Fracture Mechanics of Rock, ed. B.K. Atkinson, Academic Press, 77-9 (987). [5] Sammis, C. G. & Julian, B. R. Fracture instabilities accompanying dike intrusion. J. Geophys. Res. 9, (987). [6] Tait, S., Jaupart, C. & Vergniolle, S. Pressure, gas content and eruption periodicity of a shallow, crystallising magma chamber. Earth Planet. Sci. Lett. 9, 7- (989). [7] Kavanagh, J. L. & Sparks, R. S. J., Insigths of dyke emplacement mechanics from detailed D dyke thickness datasets. J. Geol. Soc. London 68, (). [8] Lawn, B.R., & Wilshaw, T.R. Fracture of Brittle Solids. Cambridge Univ. Press, pp. (975). [9] Schultz, R.A., Mège, D. & Diot, H. Emplacement conditions of igneous dikes in Ethiopian Traps. J. Volcanol. Geotherm. Res. 78, (8). [] Rubin, A. Tensile fracture of rock at high confining pressure implications for dike propagation. J. Geophys. Res. 98, (99). [] Saemundsson, K. Outline of the geology of Iceland. Jökull 9, 78 (979). [] Walker, G.P.L. Acid volcanic rocks in Iceland. Bull. Volcanol. 9,75- (966). [] Walker, G.P.L. The structure of eastern Iceland. In: Geodynamics of Iceland and the North Atlantic Area, ed. Kristiansson, L., Reidel, Dordrecht, pp (97). [] Walker, G.P.L. The Breiddalur Central Volcano, Eastern Iceland. J. Geol. Soc. 8, 75-9 (96). [5] Burchardt, S. & Gudmundsson, A. The infrastructure of Geitafell Volcano, Southeast Iceland. In: Studies in Volcanology: The Legacy of George Walker, eds. Thordarson, T., Self, S., Larsen, G., Rowland, S. & Hoskuldsson, A., Spec. Pub. IAVCEI, Geol. Soc. London, pp. 9-7 (9). [6] Burchardt, S. New insights into the mechanics of sill emplacement provided by field observations of the Njardvik Sill, Northeast Iceland. J. Volcanol. Geotherm. Res. 7, 8-88 (8). [7] Schmincke, H.-U. Volcanism. Springer, Berlin, pp. (6). [8] Carracedo, J. C. et al. Hotspot volcanism close to a passive continental margin: the Canary Islands. Geol. Mag. 5, 596 (998). [9] Carracedo, J.C. Growth, structure, instability and collapse of Canarian volcanoes and comparisons with Hawaiian volcanoes. J. Volcanol. Geotherm. Res. 9, -9 (999). 6

27 [] Staudigel, H., Feraud, G. & Giannerini, G. The history of the intrusive activity on the island of La Palma (Canary Islands). J. Volcanol. and Geotherm. Res. 7, 99- (986). [] Ancochea, E., et al. Constructive and destructive episodes in the building of a young Oceanic Island, La Palma, Canary Islands, and genesis of the Caldera de Taburiente. J. Volcanol. Geotherm. Res. 6, -6 (99). [] Carracedo, J. C. The Canary Islands: an example of structural control on the growth of large oceanic-island volcanoes. J. Volcanol. Geotherm. Res. 6, 5- (99). [] Walter, T.R., Troll, V.R., Cailleau, B., Belousov, A., Schmincke, H.-U., Amelung, F. & vor der Boogaard, P. Rift zone reorganization through flank instability in ocean island volcanoes: an example from Tenerife, Canary Islands. Bull. Volcanol. 67, 8-9 (5). 7