The Timor collision front represents the southernmost, youngest deformation front resulting from the ≈8 Myr-old, diachronous collision between the Australian continental plate and the oceanic domain of the Banda Sea (e.g., Charlton et al., 1991; Hall, 2009; Harris, 2011). This region marks a drastic change in both tectonic processes and nature of converging plates, shifting from an oceanic subduction along the Java Trench to a continental collision along the Timor Trough (e.g., Audley-Charles, 2004). In this region, the Timor collision front bounds a ≈100 km-wide accretionary wedge mainly composed of the northwest Australian shelf sediments. When locally exhumed and uplifted, these sediments outcrop on the Timor, Sumba, Roti and Savu islands (e.g., Roosmawati and Harris, 2009) to form the Outer Banda arc in opposition to the Inner Banda arc, which corresponds to the volcanic arc associated with the initial subduction stage (Figure 2A). The Inner Banda arc is the eastward extension of the Java-Sumatra volcanic arc (Norvick, 1979), both resulting from the large-scale convergence of the Indo-Australian plate toward Eurasia. Transitional isotopic and chemical signatures of active and extinct volcanoes along and across the arc reflect spatial and temporal variations in the nature of subducted material and yield evidence for limited to significant continental sediments influence (e.g., van Bergen et al., 1989; Varekamp et al., 1989). Radiometric dating results indicate that continental contamination likely started at least 5 Ma ago in the Wetar/Central Timor region (Elburg et al., 2005; Scotney et al., 2005), which has been interpreted, along with other lines of evidence, as a constraint for the onset of the arc-collision (Harris, 2011).

The obliquity of the Australian convergence in relation to the changing orientation of the Timor collision front entails variable degrees of shortening and strike-slip partitioning, as shown on Figure 2A. The margin-orthogonal shortening component is expected to increase from the SW/NE-striking Timor trough section (offshore Savu island to eastern tip of Timor Island, see Figure 2B) to the WSW/ESE-striking section (between East Timor and Jamdena Islands, see Figure 2C) as the orientation of the collision front becomes nearly east-west. Inversely, the margin-parallel strike-slip component is expected to be larger along the SW/NE-striking Timor trough section than that along the WSW/ESE-striking section, and should definitely become predominant over thrusting along the sub-N/S-striking section of the Tanimbar Trough, running parallel to the N12E Australian convergence vector. Below we discuss the morphology of the Roti and Jamdena regions.

Along the SW/NE-striking Roti region (Figure 2B), we identify 1) one main region with elongated, narrow folds oriented N80E on average on the northern side of the Timor Trough, 2) a discontinuous collision front, and 3) two sets of normal faults with azimuths of N75 ± 5E and N25E dissecting the western Ashmore platform surface south of the trough.

Between 122°E and 124°45′E, narrow folds are almost continuous over the 300 km-long stretch, outlining a more or less well-developed distinctive outer accretionary prism reaching up to 18 km-wide and 800 m-high (profiles R2 and R3 in Figure 3A). They likely accommodate several tens of kilometers of shortening of incoming NW shelf material at the foot of the Timor wedge. Where the folded section is shorter, the folds form a low (200 m-high) topography at the foot of the kilometric-scale structural step as if recently added to a pre-existing wedge (R2 profile offshore in Figure 3A). West and east of this folded stretch (between 121°45′E and 122° and between 124°45′E and 125°10′E), the wedge morphology changes rather abruptly, exhibiting a rugose texture similar to the morphology of the upper wedge seen above the prominent folds (Figures 2B, 3A). However, both folded and rough regions show a consistent, first-order kilometric-scale topographic step as well as a convex-upward front (Figure 3A).

Along the first western 100 km of the folded stretch, the N60E-oriented deformation front shows several, kilometric left-stepping offsets ranging between 5 and 15 km (Figures 2B, 3A), which likely accommodate the kinematically expected strike-slip component of the Australian convergence as described above. We do notice Z-shaped features east of that offset, around 123°10′E, aligned with a geographic step between the Roti and Timor coastlines (Figure 2A) and coinciding with the location of the Semau left-lateral strike-slip fault, which has been proposed to transfer most of the plate convergence from the collision front to the inner arc structures (e.g., Nugroho et al., 2009; Koulali et al., 2016).

The N75 ± 5E-oriented normal fault set is well-developed in the western part of the Roti dataset and dissects the regularly steepening Australian northwest shelf surface over more than 300 km along and 100 km away from the collision front. These normal faults are spatially correlated with the folds location (i.e., in the convergence direction) (Figures 2B, 3A). The N25E-oriented normal faults, sub-parallel to the convergence direction, are visible over a 150 km long section, east of the Semau Fault location and west of the possible offshore continuation of the Belu strike-slip fault. These faults are predominant over the N80E normal faults forming N25E “en échelon” fault arrays exhibiting strike-slip accommodation processes. Between 124°E and 124°30′E, a nearly continuous, 80 km-long en-échelon array correlates with a major shift, or left-stepping offset, of the collision front location northeastwards and with the possible offshore continuation of the Belu strike-slip fault, which might also contribute to plate convergence transfer to the Inner Banda arc structures. This significant en-échelon array marks an abrupt morphological change in the northwest shelf eastwards, followed by a smooth, nearly undisturbed morphology. Note a few narrow folds have developed in the Timor wedge in this region.

Further east, along the WSW/ESE-striking Jamdena region, we identify very similar tectonic features to those observed in the Roti region: 1) a frontal wedge region marked by elongated folds, however, here, associated with 2) a rather continuous collision front, and 3) two sets of normal faults dissecting the Sahul platform surface (Figure 2C). Between 128°20′E and 130°30′E, broad-folds oriented N95E–N70E, i.e., roughly parallel to the collision front, form a distinctive accretionary prism north of the Timor Trough (Figure 2C), nevertheless less prominent than that in the Roti region (Figure 2B). The width of the folded region reaches 19 km throughout this area (Figure 3B). As described in the Roti dataset, the folded region forms part of a kilometric-scale structural step exhibiting a steeper frontal slope compared to those offshore West Timor and Roti Islands (≈4.3° and ≈2.3°, respectively), as is observed in the incoming margin slope (Figure 3B). West of this folded region, the Timor wedge morphology changes abruptly to a rather heterogeneous morphology, marked by a rugged texture (particularly visible between 127°40′E and 128°25′E up to the front) as previously described in the Roti region. The nearly continuous deformation front does not feature any clear left-stepping offsets here, as expected in relation to the changing orientation of the collision front (Figure 2A). Two sets of normal faults are visible on the incoming Australian margin with comparable azimuths to those observed in the Roti region (N75 ± 10E and N5E, respectively). However, their seafloor imprint is more subdued here and the N5E-oriented en-échelon fault array is very restricted. Note here again a spatial correlation between the normal faults developed in the northwest shelf and the outer folds of the Timor wedge.

To constrain the nature and geometry of underlying fault systems contributing to the variable morphologies described in the Timor wedge and in the Australian northwest shelf, we interpret and correlate five high-resolution 2D seismic reflection profiles acquired offshore East Timor and Leti Islands (i.e., in Jamdena region, Figures 2, 3) (Figures 2, 3). The five profiles (MH07-02, MH07-04, MH07-11, MH07-16, and MH07-22, dubbed profiles 2, 4, 11, 16, and 22, respectively, in the following) image both the Australian northwest shelf and the Timor wedge between 128°E and 129°20′E, over spatial lengths ranging between 72 and 150 km, and record lengths of 8–12 s two-way travel time (TWT).

We map a series of prominent sedimentary reflectors on the northwest Australian shelf (Figures 4A,B) that we correlate with a few published known horizons identified in this area (e.g., Keep et al., 2002 ; Baillie et al., 2004, see Supplementary Figure S1) with observed reflectors analyzed here to delimit the Cenozoic sedimentary section (Figures 4A,B). In the Timor wedge, we map three main seismic units : 1) an undeformed, thick bright unit (300 ms TWT on average, see Figures 4A,B) likely indicating pelagic or semi-pelagic sediments (e.g., Breen et al., 1986) on top of a 2) a chaotic reflection of likely intensively deformed old sediments representing the upper wedge of the rugged regions, and 3) a series of parallel reflectors characteristic of the Cenozoic sedimentary section either observed continuously beneath the rigid upper wedge (e.g., profile 4, Figures 4A,) or intensively stacked atop one another at the front of the Timor wedge (e.g., Figures 4B,Bii). We finally map normal and thrust faults from the edge of the Sahul platform to the upper Timor wedge (Figures 4A,B).

The northwest Australian shelf exhibits clear, flat and parallel reflectors changing from being sub-horizontal on the Sahul platform to north-dipping in the downgoing northwest shelf slope (Figure 5). The overall thickness of this Cenozoic sedimentary sequence at the collision front increases significantly from the western profiles to the eastern ones (i.e., reaches about 1.5 s TWT on profile 2 vs. 3 s TWT along profile 22, Figures 5, 6). Most of the reflectors are regularly disturbed by sets of conjugate normal faults, extending to the lowermost strong reflection (likely being the top of the Triassic unit, Keep et al., 2002). The deeper and older units show constant thickness and fault offsets attesting pre-tectonic deposition while the upper units (i.e., younger than Pliocene) exhibit syn-tectonic deposition as recorded by larger deposition centers along and between antithetic normal faults indicating a recent increase in fault activity (Figures 4, 5). Some of these normal faults locally outcrop at the seafloor forming the ≈N75E-striking normal faults we have described above, particularly noticeable along profile 22.

Toward the Timor wedge, the downgoing Cenozoic section on profiles 2 and 4 shows mostly landward-verging normal faults over the entire profile lengths. The abrupt change of seismic reflection signal between the parallel reflectors and the chaotic wedge unit (Figure 4A) indicates the location of a single, shallow-dipping frontal decollement extending for at least 20 km north of the deformation front (Figures 4, 5). This thrust bounds both the base of the consolidated, inactive upper wedge, where no clear structure nor recent folding can be seen, and on top of which the thick bright pelagic cover extends over the entire wedge width. The profile 11 shows a similar wedge architecture to profiles 2 and 4 with an additional sedimentary section at the foot of the accretionary wedge, involving the youngest Cenozoic sedimentary units as well as the Trough sediments (Figure 5). On profiles 16 and 22 (Figure 5), the thick incoming Cenozoic sequence displays both landward and seaward-verging normal faults. The latter show reverse reactivation close to the front. Differing from profiles 2 and 4 (Figure 5), these sediments are not observed under the Timor wedge but observed as part of it, displaying thin wedge slices stacked atop one another along long, steep thrusts rooted on a relatively deep horizontal basal decollement. They form rather unconsolidated, weaker, recent folded morphology described earlier, in opposition to the old rigid upper wedge described on profiles 2 and 4 (Figure 5). The outer wedge in this section is not topped by the undeformed pelagic (or semi-pelagic) sediment layer described above. That sedimentary layer actually stops where the first, inner, steep slice of Cenozoic sediments starts being accreted to the old upper wedge (see Figures 4B, 5).

Our morphological, topographic, and seismic data analyses reveal two distinct tectonic regimes varying from total underthrusting to total accretion, which are seemingly controlled by inherited structures of the northwest shelf in the Jamdena region (Figures 5, 6A). Based on similar morphological characteristics highlighted both on the northwest Australian shelf and in the Timor wedge, we assume that comparable tectonic wedge architectures and tectonic regimes lie in the Roti region as well, for which we do not have any seismic constraints. The underthrusting regime is characterized by 1) the total underthrusting of a rather thin section of incoming Cenozoic sediments (Figures 6B,C), exhibiting mostly landward-verging normal faults, 2) a single, low-dipping shallow decollement (Figures 6B,C), and 3) a consolidated, rigid, and non-recently deformed upper wedge (see also profiles 2 and 4). The accretion regime is characterized by 1) the total accretion of a rather thick section of the incoming Cenozoic sediments (Figures 6B,C), exhibiting more seaward-verging normal faults, 2) a deep frontal decollement with steep frontal thrusts (Figures 6B,C), and 3) the development of an unconsolidated lower accretionary wedge in front of the older wedge (profiles 16 and 22).

Close to the deformation front, some of the seaward-verging normal faults cutting the entire sediment thickness show thrust reactivation (profile 22, Figure 5), indicating that the growth of the recent accretionary prism likely uses pre-existing, favourably oriented normal faults of the incoming northwest shelf sediments. The spatial distribution and similar orientation of both the N75E faults developed in the northwest shelf and the narrow folds of the accretionary prism along the collision front in the Jamdena region (Figures 2, 3) support this hypothesis. It is also possible that these prevailing tectonic regimes give evidence for different, but concurrent, chronological stages of deformation along the collision front where the Timor wedge evolves from a nascent collision stage to a more mature stage as it incorporates a developing accretionary prism. This would be in keeping with the larger time and space scale diachronous growth and exhumation of the Banda arc islands, influenced by the shape (e.g., Scott plateau protrusion) and sedimentary structure of the incoming Australian shelf (Keep et al., 2002; Harris, 2011).

The morphological and structural transition from the underthrusting to the accretion regime is associated with a decollement depth increase (from 2 s TWT to almost 4 s TWT, Figure 6C) thus forming a structural step that one could consider a lateral ramp (Tréhu et al., 2019). If significant enough, lateral ramps can act as barriers to slip propagation (e.g., Qiu et al., 2016—Himalayan range). Morpho-structural transitions can therefore be used to infer fault segmentation along the Timor Trough, which indirectly helps constraining geohazards—earthquake magnitude being directly related to the rupture area.

Along the Timor wedge, we identify four morphological transition zones (located around 122°E, 125°10′E, and 128°20′E, and 131°E along the collision front, Figure 7), likely indicating tectonic regimes shifts and therefore structural variations (i.e., lateral ramps). We characterize four fault segments, which alternate between underthrusting and accretion dominated processes along most of the Timor Trough segment of the Outer Banda arc (Figure 7). The two accretion segments A1 and A2 are 350 km and 250 km-long, respectively. The lengths of U1 and U2 underthrusting segments are 200 and 300 km, respectively, which are maximum values since we do neither have bathymetric nor 2D seismic data in those fault segments to verify the absence of other structural complexities constraining smaller segments. The two shallow aseismic underthrusting segments U1 and U2 spatially correlate with two relatively quiet seismic zones of the intermediate seismicity as shown on Figure 7 (ISC catalog : 1964–1996, relocated hypocenters shown for earthquakes deeper than 50 km, Das, 2004 – personal communication, Figure 7), including the Wetar Seismic Gap identified as a slab window (Sandiford, 2008; Ely and Sandiford, 2010) (Figure 7). This spatial correlation between shallow and deeper processes suggests that the present-day collision front segmentation we propose here reflects the large-scale and long-term geodynamics of the collision between the northwest Australian shelf and the Outer Banda arc.

In relation to the differences in the fault geometry, thicknesses of underthrusted sediments, and overall tectonic structure of the Timor wedge, the underthrusting and accretion segments are likely experiencing different seismic behaviors. In the following, we refer to subduction examples rather than collision ones in an attempt to constrain those behaviors. In the absence of local geodetic and geochemical investigations to directly constrain the interface seismic state, we do take advantage of intensive investigations published on analogous oceanic convergent settings (e.g., Moore et al., 2007; Kodaira et al., 2012; Han et al., 2017; Kodaira et al., 2017; Li et al., 2018; Tréhu et al., 2019; Von Huene et al., 2019; Olsen et al., 2020; Watson et al., 2020).

In the accretion segments, the lack of high-fluid content, young Australian Cenozoic sediments under the wedge indicates that the basal thrust would be relatively strong (i.e., with a relatively high-seismic coupling) and therefore not that prone to regular great earthquakes nucleation at depth, but likely regularly rupturing with smaller magnitude earthquakes (Figure 7; Li et al., 2018). This would be consistent with the moderate (M < 6.5) seismicity recorded in both Jamdena and Roti regions. The largest of these moderate magnitude earthquakes–with magnitudes of M_w 6.2 and M_w 5.7—successively occurred offshore west Timor Island on August 28, 2018 (Figure 7; Supplementary Figure S2) on a 68° north-dipping plane at 14 km depth and on a 33° north-dipping plane at 8 km depth, respectively (Supplementary Table S1). Epicentre locations and focal mechanisms from the USGS catalog suggest that the larger one likely reactivated one of the relatively steep, pre-existing, favourably oriented normal fault as a thrust fault (see Supplementary Figure S2), and that the smaller one, ruptured the most frontal thrust of the recent Timor wedge. Slump traces mapped in the vicinity of those faults (Figure 2B) might result from shaking during those two earthquakes and could therefore be good indicators of recent earthquakes in the region. In combination with our morpho-tectonic and structural observations, we therefore envision three types of active fault sources for the accretion segments : 1) the most frontal thrust of the Timor wedge, 2) the incoming, thrust-reactivated normal faults, and 3) the backstop thrust (or megasplay), which represents a mechanical boundary between the old, stronger wedge and the newly, weaker and fluid-rich segment (Moore et al., 2007) (Figure 7, line AA’). Because of their relative steepness (note that we do not derive any dip measurement from the time sections), slip along these faults could result in significant seafloor uplift and subsequent tsunami waves even for moderate earthquakes (Singh et al., 2011).

In the underthrusting segments, the overall high-fluid content of a relatively thick, young Australian Cenozoic sediment sequence passing beneath the rigid upper wedge, would likely generate elevated pore pressure on the basal decollement, resulting in the weakening of thrust (Han et al., 2017) (Figure 6). The absence of deformation in the overlying wedge supports the possibility of a weak basal thrust (e.g., Tréhu et al., 2019). Such a mechanical weakening would favor stable sliding at shallow depths (Li et al., 2018), which, in the Timor region, possibly accounts for the present-day, nearly aseismic behavior of the Banda arc collision front (Figures 1, 7). However, progressive dewatering and consolidation of the Cenozoic sediments at greater depths (i.e., at elevated temperature and pressure) could result in the formation of a large, uniform locked asperity where nucleation of regular and great earthquakes could take place (Ruff, 1989; Li et al., 2015; Li et al., 2018) (Figure 7, line BB’). Such a rupture could either be restricted to the deeper portion of the basal thrust or extend up to the front. In the first case scenario, the stable sliding shallow area, seemingly weakly coupled, might impede updip co-seismic slip propagation as characterized for the M_w 7.8, 2016, Pedernales (Ecuador) and the M_w 7.6, 2012, Costa Rica earthquake rupture patterns (Dixon et al., 2014; Rolandone et al., 2018, respectively). In the latter case, that same conditionally stable shallow basal thrust area could become unstable and prone to rupture under certain circumstances (Scholz and Campos, 2012). This process possibly explains the unexpected trench-rupturing M_w 9.0, 2011, Tohoku Oki earthquake Japan, 2011 (Kodaira et al., 2012) and the M_w 8.4, 2010, Maule earthquake (Tréhu et al., 2019). The main active fault of the 200–300 km length consisting of underthrusting segments with locked basal decollement, which accommodates most of the convergence, and could possibly generate large magnitude earthquakes and tsunamis (Figure 7, line BB’).