3D Tomographic Imaging of P Wave Velocity Structure Beneath Java Island using Fast Marching Tomography Method

 Java Island one of the best locations for geophysics research, as it’s located near the edge of the junction of continental Eurasia Plate and the Indo-Australia Plate with the movement of the plates 6 cm/year. Its needed the accurate imaging of subsurface structure to understanding complex tectonic setting. The first arrival time from local earthquake of Java (6,360-9,150 S and 105,890-115,540 E) with magnitude greater than 4,5 Mw occurred from 2011 until 2013 recorded by local network seismograph has been inverted for three dimensional variation of the depth to the P wave velocity in the Java island. At the same time, earthquake hypocenter location has been corrected simultaneous. This research apply a new tomographic inversion scheme FMTOMO that has been developed by Rawlinson (2004). The results of inversion show that there are three layer in 100 km from the surface at continental plate side (Eurasia). The continental upper crust has P wave velocity variation about 4-5,5 km/s in the depth until 30 km from the surface. There are low velocity anomaly inclined towards the slab which probably have associated with shallow earthquake along the

I. INTRODUCTION 1 he subduction zones on the crust of the Earth is related with the seismic and volcanic activity. Java island, located at the junction of two major tectonic plates (Indo-Australian plate and Eurasian plate), is formed by a complex mosaic of oceanic plates, island arcs, continental shelf and crustal fragments [1]. Tectonic stucture of Java island is dominated by the almost orthogonal convergence between the Australian plate and Eurasia (67-70 mm/year, today).
The Indo-Australia plate is pushing the Eurasia plate. In north of Java island, Indo-Australia plate is moving 100-200 under the Eurasian plate [2]. The deeper slab beneath Java, appearing as an inclined high-velocity feature, seems to be continuous from the Earth's surface to the lower mantle [3].
To understand the mechanisms of the process of tectonic activity, seismicity and volcanism, we need some tools. Seismic tomography is one of the most powerful tools to obtain these constraints [4]. This is a form of seismic traveltime inversion that is used to constrain 2D and 3D models of the Earth represented by a significant number of parameters [5]. Actually the earthquake cause the propagate of seismic wave that bring some information about interior structure of the earth. Seismic data in seismograph represent one of the most valuable resources and important information for investigating the internal structure and composition of the earth. 1 Uswatun Chasanah and Bagus Jaya Santosa are with Department of Physics, Institut Teknologi Sepuluh Nopember, Surabaya, 60111, Indonesia.
E-mail: chasanah14@mhs.physics.its.ac.id; uswah23@gmail.com; bjs@physics.its.ac.id The accurate of hypocentre location is one of the main things when studying the seismicity of the area within the parameters of the earthquake because there is important information from ongoing seismotectonic process [6]. The earthquake parameters are usually magnitude, origin time, epicentre and depth. The accuracy of seismic parameter calculation is highly dependent on the structure of seismic wave velocity models used. The aim of this local scale studies on this paper is to explore the deep structure of slabs beneath Java Island by means of tomographic imaging using P wave phases.

II. METHOD
The 431 P wave phases was picked from waveform 3 components that recorded by 35 local network seismograph. And it was inverted using FMTOMO that calculated using fast marching tomography method. This processes obtaining the value of 3D velocity to determine the real structure of tectonic settings under the Java island.
Origin time value of P waves is obtained by picking the seismic waveform from each station. In this study, picking process of arrival time of P wave is using software Seisgram2K ( Figure. 1). FMTOMO uses the fm3d package to solve the forward problem. Fm3d is based on the multi-stage Fast Marching Method (FMM), a grid based eikonal solver (1) that implicitly tracks the evolution of waveform in 3-D layered media. For more details, refer to Sethian (1996), Sethian& Popovici (1999), Sethian (1999) and Popovici & Sethian (2002) [7], [8]. The eikonal equation states that the magnitude of the traveltime gradient at any point along a wave front is equal to the inverse of the wavespeed at that point [9] and may be written as: where ∇x is the gradient operator, T is traveltime and s(x) is slowness [10]. If we represent some elastic property of the subsurface (e.g. velocity) by a set of model parameters m, then a set of data (e.g. traveltimes) d can be predicted for a given sourcereceiver array by line integration through the model. The inversion step, which involves the adjustment of the model parameters m to better satisfy the observed data dobs through the known relationship.
= (2) In travel time tomography, the functional G is non-linear because the ray path depends on the velocity structure [11]. For an initial estimate m0 of the model parameters, comparing this equation (3) = 0 (3) with the observed travel time dobs gives an indication of the accuracy of the model. The misfit can be quantified by constructing an objective function S(m)(4), consisting of a weighted sum of data misfit and regularization terms, that is to be minimized. The objective function is given (4).  Table 3. The ak135 velocity model was used for the reference, or starting, velocity model. Damping (ϵ) and smoothing (η) values were chosen to solve variations in deep mantle and crustal structure may account for some of the data misfit. We used a 20-dimensional subspace scheme on six iterations to solve both P inversions. However, in some cases the subspace was reduced by Singular Value Decomposition (SVD) orthogonalization to remove unnecessary basis vectors during the inversion step. Iterations are repeated until the contribution of the next cycle becomes negligible. In case of our study, six iterations were enough to achieve sufficient convergence [12].    The earthquake hypocentre is located just below this level (~60 km depth). This implies that it is located in the rigid crust that is the most probable location for stress accumulation.

IV. CONCLUSION
We have used a new iterative non-linear tomography scheme to simultaneously invert seismic source datasets for the 3-D seismic structure of the Java island lithosphere.
The deployment of a exceptionally dense temporary seismic network allowed the determination of the crustal structure of Java island at high resolution.
The result of inversion show that there are three layer in 100 km from the surface at continental plate side (Eurasia).
The upper crust have Vp about 4-5,5 km/s. The lower crust have Vp 6,5-7 km/s. Upper mantle have Vp higher value, about 6,5-8,5 km/s. The oceanic plate (Indo-Australia) have two layer in 100 km from the surface. The oceanic crust have Vp about 6,5-7 km/s. The oceanic mantle have Vp 6,5-8,2 km/s as describe in Table 1 and Table 2. There are low velocity anomaly inclined towards the slab which probably have associated with shallow earthquake along the slab.