Marine terraces of the last interglacial period along the Pacific 1 coast of South America (1°N-40°S)

. Tectonically active coasts are dynamic environments characterized by the presence of multiple marine 9 terraces formed by the combined effects of wave-erosion, tectonic uplift, and sea-level oscillations at glacial-cycle 10 timescales. Well-preserved erosional terraces from the last interglacial sea-level highstand are ideal marker horizons 11 for reconstructing past sea-level positions and calculating vertical displacement rates. We carried out an almost 12 continuous mapping of the last interglacial marine terrace along ~5,000 km of the western coast of South America 13 between 1°N and 40°S. We used quantitatively replicable approaches constrained by published terrace-age estimates 14 to ultimately compare elevations and patterns of uplifted terraces with tectonic and climatic parameters in order to 15 evaluate the controlling mechanisms for the formation and preservation of marine terraces, and crustal deformation. 16 Uncertainties were estimated on the basis of measurement errors and the distance from referencing points. Overall, 17 our results indicate a median elevation of 30.1 m, which would imply a median uplift rate of 0.22 m/ka averaged over 18 the past ~125 ka. The patterns of terrace elevation and uplift rate display high-amplitude (~100–200 m) and long- 19 wavelength (~10 2 km) structures at the Manta Peninsula (Ecuador), the San Juan de Marcona area (central Peru), and 20 the Arauco Peninsula (south-central Chile). Medium-wavelength structures occur at the Mejillones Peninsula and 21 Topocalma in Chile, while short-wavelength

and Carranza, Chile. We interpret the long-wavelength deformation to be controlled by deep-seated processes at the 23 plate interface such as the subduction of major bathymetric anomalies like the Nazca and Carnegie ridges. In contrast, 24 short-wavelength deformation may be primarily controlled by sources in the upper plate such as crustal faulting, 25 which, however, may also be associated with the subduction of topographically less pronounced bathymetric 26 anomalies. Latitudinal differences in climate additionally control the formation and preservation of marine terraces. 27 Based on our synopsis we propose that increasing wave height and tidal range result in enhanced erosion and 28 morphologically well-defined marine terraces in south-central Chile. Our study emphasizes the importance of using 29 systematic measurements and uniform, quantitative methodologies to characterize and correctly interpret marine 30 terraces at regional scales, especially if they are used to unravel tectonic and climatic forcing mechanisms of their 31 formation. This database is an integral part of the World Atlas of Last Interglacial Shorelines (WALIS), published 32 online at http://doi.org/10.5281/zenodo.4309748 (Freisleben et al., 2020). 33

Introduction 34
Tectonically active coasts are highly dynamic geomorphic environments and they host densely-populated centers and 35 associated infrastructure (Melet et al., 2020). Coastal areas have been episodically affected by the effects of sea-level 36 changes at glacial timescales, modifying the landscape and leaving behind fossil geomorphic markers, such as former 37 paleo-shorelines, and marine terraces (Lajoie, 1986). One of the most prominent coastal landforms are marine terraces 38 that were generated during the protracted last interglacial sea-level highstand that occurred ~125 ka ago (Siddall et 39 al., 2006;Hearty et al., 2007;Pedoja et al., 2011). These terraces are characterized by a higher preservation potential, 40 which facilitates their recognition, mapping, and lateral correlation. Furthermore, because of their high degree of 41 preservation and relatively young age, they have been used to estimate vertical deformation rates at local and regional 42 scales. The relative abundance and geomorphic characteristics of the last interglacial marine terraces make them ideal 43 geomorphic markers with which to reconstruct past sea-level positions and to enable comparisons between distant 44 sites under different climatic and tectonic settings. 45 The Western South American Coast (WSAC) is a tectonically active region that has been repeatedly affected by 46 megathrust earthquakes and associated surface deformation (Beck et al., 1998 of reliable data points has revealed a need to re-examine the last interglacial marine terraces along the WSAC based 56 on standardized methodologies in order to obtain a systematic and continuous record of marine terrace elevations 57 along the coast. This information is crucial in order to increase our knowledge of the climatic and tectonic forcing 58 mechanisms that contributed to the formation and degradation of marine terraces in this region. 59 Marine terrace sequences at tectonically active coasts are landforms formed by wave erosion and/or accumulation of 60 sediments resulting from the interaction between tectonic uplift and superposed oscillating sea-level changes (Lajoie,61 The analysis of elevation patterns based on shoreline-angle measurements at subduction margins has been largely used 68 to estimate vertical deformation rates and the mechanisms controlling deformation, including the interaction of the 69 upper plate with bathymetric anomalies, the activity of crustal faults in the upper plate, and deep-seated processes 70 such as basal accretion of subducted trench sediments (Taylor et al., 1987;Hsu, 1992;Macharé and Ortlieb, 1992 represents a 1D descriptor of the marine terrace elevation, whose measurements are reproducible when using 73 quantitative morphometric approaches (Jara-Muñoz et al., 2016). Furthermore, the estimation of the marine terrace 74 elevations based on shoreline angles can be further improved by quantifying their relationship with paleo-sea level, 75 also known as the indicative meaning (Lorscheid and Rovere, 2019). 76 In this continental-scale compilation of marine terrace elevations along the WSAC, we present systematically mapped 77 shoreline angles of marine terraces of the last (Eem/Sangamon) interglacial obtained along 5,000 km of coastline 78 between 1°N and 40°S. In this synthesis we rely on chronological constraints from previous regional studies and 79 compilations (Pedoja et al., 2011). For the first time we are able to introduce an almost continuous pattern of terrace 80 elevation and coastal uplift rates at a spatial scale of 10 3 km along the WSAC. Furthermore, in our database we 81 compare tectonic and climatic parameters to elucidate the mechanisms controlling the formation and preservation of 82 marine terraces, and patterns of crustal deformation along the coast. This study was thus primarily intended to provide 83 a comprehensive, standardized database and description of last interglacial marine terrace elevations along the 84 tectonically active coast of South America. This database therefore affords future research into coastal environments 85 to decipher potential tectonic forcings with regard to the deformation and seismotectonic segmentation of the forearc; 86 as such this database will ultimately help to decipher the relationship between upper-plate deformation, vertical motion 87 and bathymetric anomalies and aid in the identification of regional fault motions along pre-existing anisotropies in the 88  (Garreaud, 2009). Martinod et al. (2016b) proposed that latitudinal 281 differences in climate largely influence coastal morphology, specifically the formation of high coastal scarps that 282 prevent the development of extensive marine terrace sequences. However, the details of this relationship have not 283 been conclusively studied along the full extent of the Pacific coast of South America. 284

Methods 285
We combinedand describe in detail belowbibliographic information, different topographic data sets, and uniform 286 morphometric and statistical approaches to assess the elevation of marine terraces and accompanying vertical 287 deformation rates along the western South American margin. 288

3.1.
Mapping of marine terraces 289 Marine terraces are primarily described based on their elevation, which is essential for determining vertical 290 deformation rates. The measurements of the marine terrace elevations of the last interglacial were performed using 291 TanDEM-X topography (12 and 30 m horizontal resolution) (German Aerospace Center (DLR), 2018), and digital 292 terrain models from LiDAR (1, 2.5, and 5 m horizontal resolution). The DEMs were converted to orthometric heights 293 by subtracting the EGM2008 geoid and projected in UTM using the World Geodetic System (WGS1984) using zone 294 19S for Chile, zone 18S for southern/central Peru, and zone 17S for northern Peru/Ecuador. 295 To trace the MIS-5 shoreline, we mapped its inner edge along the west coast of South America based on slope changes 296 on TanDEM-X topography at the foot of paleo-cliffs (Jara-Muñoz et al., 2016) ( Fig. 2A  and Jara-Muñoz et al. (2015) for the area between 34° and 38°S. We define the term "referencing point" for these 301 previously published terrace heights and age constraints. The referencing point with the shortest distance to the 302 location of our measurements served as a topographical and chronological benchmark for mapping the MIS-5 terrace 303 in the respective areas. In addition, this distance is used to assign a quality rating to our measurements. 304 In addition to MIS-5e, we also mapped MIS-5c in areas with high uplift rates such as at the Manta Peninsula, San 305 Juan de Marcona, Topocalma, Carranza, and Arauco. Although we observed a terrace level correlated to MIS-5a in 306 the Marcona area, we excluded this level from the database due to its limited preservation at other locations and lack 307 of chronological constraints. Our assignment of mapped terrace levels to MIS-5c is primarily based on age constraints 308 However, in order to evaluate the possibility that our correlation with MIS-5c is flawed, we estimated uplift rates for 310 the lower terraces by assigning them tentatively to either MIS-5a or MIS-5c. We interpolated the uplift rates derived 311 from the MIS-5e level at the sites of the lower terraces and compared the differences ( Figure 3A). If we infer that 312 uplift rates were constant in time at each site throughout the three MIS-5 substages, the comparison suggests these 313 lower terrace levels correspond to MIS-5c because of the smaller difference in uplift rate, rather than to MIS-5a (Figure 314 www.terracem.com. We placed swath profiles of variable width perpendicular to the previously mapped inner edge, 322 which were used by the TerraceM algorithm to extract maximum elevations to avoid fluvial incision ( Fig. 2A and B). 323 For the placement of the swath profiles we tried to capture a local representation of marine terrace topography with a 324 sufficiently long, planar paleo-platform, and a sufficiently high paleo-cliff, simultaneously avoiding topographic 325 disturbance, such as colluvial wedges or areas characterized by river incision.
The four parameters that we included in our quality rating (QR) comprise a) the distance to the nearest referencing reliability of the dating method, but can be increased by good age constraints of adjacent terrace levels or detailed 422 morphostratigraphic correlations, such as in Chala Bay ( Fig. 2A) (Goy et al., 1992;Saillard, 2008). We further used 423 this confidence value to quantify the quality of the age constraints in the WALIS template. 424 To account for the different uncertainties of the individual parameters in the QR, we combined and weighted the 425 parameters DRP and CRP in a first equation claiming 60% of the final QR, ET in a second and R in a third equation 426 weighted 30% and 10%, respectively. We justify these percentages by the fact that the distance and confidence to the 427 nearest referencing point is of utmost importance for identifying the MIS-5e terrace level. The measurement error 428 represents how well the mapping of the paleo-platform and paleo-cliff resulted in the shoreline-angle measurement, 429 while the topographic resolution of the underlying DEM only influences the precise representation of the actual 430 topography and has little impact on the measurement itself. The coefficient assigned to the topographic resolution is 431 multiplied by a factor of 1.2 in order to maintain the possibility of a maximum QR for a DEM resolution of 5 m. 432 Furthermore, we added an exponent to the first part of the equation to reinforce low confidence and/or high distance 433 of the referencing point for low quality ratings. The exponent adjusts the QR according to the distribution of distances 434 from referencing points, which follows an exponential relationship (Fig. 4D). 435 The influence of each parameter to the quality rating can be observed in Fig. 4. We observe that for high DRP values 436 the QR becomes constant; likewise, the influence of QR parameters becomes significant for QR values higher than 3. 437 We justify the constancy of the QR for high DRP values (> 300 km) by the fact that most terrace measurements have 438 DRP values below 200 km (Fig. 4D). The quality rating is then used as a descriptor of the confidence of marine terrace-439 elevation measurements. 440  where ΔH is the relative sea level, HSL is the sea-level altitude of the interglacial maximum, HT is the shoreline-angle 449 elevation of the marine terrace, and T its associated age (Lajoie, 1986). GEBCO Bathymetric Compilation Group, 2020) (Fig. 1). 473 To evaluate the potential correlations between tectonic parameters and marine terraces, we analyzed the latitudinal 474 variability of these parameters projected along a curved "simple profile" and a 300-km-wide "swath profile" following 475 the trace of the trench. We used simple profiles for visualizing 2D data sets; for instance, to compare crustal faults

Statistical analysis 649
Our statistical analysis of mapped shoreline-angle elevations resulted in a maximum kernel density at 28.96 m with a 650 95% confidence interval from 18.59 m to 67.85 m (2σ) for the MIS-5e terrace level (Fig. 10A). The MIS-5c terrace 651 yielded in a maximum kernel density at a higher elevation of 37.20 m with 2σ ranging from 24.50 m to 63.92 m. It is 652 important to note that the number of MIS-5c measurements is neither as high nor as continuous as compared to that 653 of the MIS-5e level. MIS-5c data points were measured almost exclusively in sites where MIS-5e reach high elevations 654 (e.g., San Juan de Marcona with MIS-5e elevations between 40 and 110 m). 655 The distribution of measurement errors was studied using probability kernel-density plots for each topographic 656 resolution (1-5 m LIDAR, 12 m TanDEM-X, and 30 m TanDEM-X). The three data sets display similar distributions 657 and maximum likelihood probabilities (MLP); for instance, LiDAR data show a MLP of 0.93 m, the 12 m TanDEM-658 X a MLP of 1.16 m, and 30 m TanDEM-X a MLP of 0.91 m (Fig. 10B). We observe the lowest errors from the 30 m 659 TanDEM-X, slightly higher errors from the 1-5 m LiDAR data, and the highest errors from the 12 m TanDEM-X. 660 This observation is counterintuitive as we would expect lower errors for topographic data sets with higher resolution 661 (1-5 m LiDAR). The reason for these errors is probably related to the higher number of measurements using the 12 m 662 TanDEM-X (1564) in comparison with the measurements using 30 m TanDEM-X (50), which result in a higher 663 dispersion and a more realistic representation of the measurement errors (Fig. 10B). In addition, the relation between 664 terrace elevations and error estimates shows that comparatively higher errors are associated with higher terrace 665 elevations, although the sparse point density of high terrace-elevation measurements prevents recognition of a clear 666 correlation from being recognized (Fig. 10C). 667 In this study we generated a systematic database of last interglacial marine terrace elevations with unprecedented 684 resolution based on an almost continuous mapping of ~2,000 measurements along 5,000 km of the WSAC. This opens 685 up several possibilities for future applications in which this database can be used; for example, marine terraces are 686 excellent strain markers that can be used in studies on deformation processes at regional scale, and thus the synthesis 687 allows comparisons between deformation rates at different temporal scales in different sectors of the margin or 688 analyses linking specific climate-driven and tectonic coastal processes, and landscape evolution. However, there are 689 a number of limitations and potential uncertainties that can limit the use of this database in such studies without taking 690 several caveats into consideration. 691 One of the most critical limitations of using the database is associated with the referencing points used to tie our 692 marine terrace measurements, which are in turn based on the results and chronological constraints provided by 693 previous studies. The referencing points are heterogeneously distributed along the WSAC, resulting in some cases of 694 up to 600 km distance to the nearest constrained point, such as in Central Peru [e.g. Pe2]. This may have a strong 695 influence on the confidence in the measurement of the marine terrace elevation at these sites. In addition, the 696 geochronological control of some of the referencing points may be based on dating methods with pronounced 697 uncertainties (e.g., amino acid racemization, electron spin resonance, terrestrial cosmogenic radionuclides), which 698 may result in equivocal interpretations and chronologies of marine terrace levels. In order to address these potential 699 factors of uncertainty we defined a quality rating (see section 3.1.), which allows classifying our mapping results based 700 on their confidence and reliability. Therefore, by considering measurements above a defined quality it is possible to 701 increase the level of confidence for future studies using this database; however, this might result in a decrease of the 702 number of measurements available for analysis and comparison. 703

Tectonic and climatic controls on the elevation and morphology of marine terraces along the WSAC 704
In this section we provide a brief synthesis of our data set and its implications for coastal processes and overall 705 landscape evolution influenced by a combination of tectonic and climatic forcing factors. This synthesis emphasizes 706 the significance of our comprehensive data set for a variety of coastal research problems that were briefly introduced 707 in section 5.1. Our detailed measurements of marine terraces along the WSAC reveal variable elevations and a 708 heterogeneous distribution of uplift rates associated with patterns of short-, medium-, and long-wavelengths. In 709 addition, we observe different degrees of development of marine terraces along the margin expressed in variable 710 shoreline-angle density. There are several possible causes for this variability, which we explore by comparing terrace-711 elevation patterns with different climatic and tectonic parameters. 712

Tectonic controls on coastal uplift rates 713
The spatial distribution of the MIS-5 marine terrace elevations along the convergent South American margin has 714 revealed several high-amplitude and long-wavelength changes with respect to tectonically controlled topography. The subduction of bathymetric anomalies has been shown to exert a substantial influence on upper-plate deformation 720 (Fryer and Smoot, 1985;Taylor et al., 1987;Macharé and Ortlieb, 1992 (Fig. 11A). 734 In contrast, small-scale bathymetric anomalies correlate in part with the presence of crustal faults perpendicular to the 735 coastal margin near, for instance, the Juan Fernandez, Taltal, and Copiapó ridges (Fig. 11B); this results in short-736 wavelength structures and a more localized altitudinal differentiation of uplifted terraces. This emphasizes also the 737 importance of last interglacial marine terraces as strain markers with respect to currently active faults, which might be 738 compared in the future with short-term deformation estimates from GPS or the earthquake catalog. In summary, short-739 wavelength structures in the coastal realms of western South America may be associated with faults that root at shallow

Climatic controls on the formation and preservation of last interglacial marine terraces 750
The latitudinal climate differences that characterize the western margin of South America may also control coastal 751 morphology and the generation and preservation of marine terraces (Martinod et al., 2016b). In order to evaluate the 752 influence of climate in the generation and/or degradation of marine terraces, we compared the number of marine 753 terrace measurements, which is a proxy for the degree of marine terrace preservation, and climatically controlled 754 parameters such as wave height, tidal range, coastline orientation, and the amount of precipitation. 755 The maximum wave height along the WSAC decreases northward from ~8 to ~2 m (see section 3.3, Fig. 11D). 756 Similarly, the tidal range decreases progressively northward from 2 to 1 m between Valdivia and San Juan de Marcona, 757 followed by a rapid increase to 4 m between San Juan de Marcona and the Manta Peninsula. We observe an apparent 758 correlation between the number of measurements and the tidal range in the north, between Illescas and Manta (Fig.  759 11F). Likewise, the increasing trend in the number of measurements southward matches with the increase in wave 760 height (Fig. 11D). An increase of wave height and tidal range may lead to enhanced erosion and morphologically well-761 expressed marine terraces (Anderson et al., 1999;Trenhaile, 2002), which is consequently reflected in a higher number 762 of measurements that can be carried out. Furthermore, we observe low values for the reference water level (< 0.7 m) 763 resulting from tide and wave-height estimations in IMCalc (Lorscheid and Rovere, 2019), which are used to correct 764 our shoreline-angle measurements in the WALIS database (see section 3.3.). 765 The control of wave-erosion processes on the morphological expression of marine terraces may be counteracted by 766 erosional processes such as river incision. We note that the high number of preserved marine terraces between 767 Mejillones and Valparaíso decreases southward, which coincides with a sharp increase in mean annual precipitation 768 from 10 to 1000 mm/yr ( Fig. 11E and F) and fluvial dissection. However, in the area with a high number of 769 measurements between the Illescas Peninsula and Manta we observe an opposite correlation: higher rainfall associated 770 with an increase of marine terrace preservation (Fig. 11E). This suggests that the interplay between marine terrace 771 generation and degradation processes apparently buffer each other, resulting in different responses under different 772 climatic conditions and coastal settings. 773 The higher greater number of marine terraces between Mejillones and Valparaíso and north of Illescas corresponds 774 with a SSW-NNE orientation of the coastline (azimuth between 200 and 220°). In contrast, NW-SE to N-S oriented 775 coastlines (azimuth between 125 and 180°), such as between the Arica and Huancabamba bends, correlate with a 776 lower number of marine terrace measurements ( Fig. 11E and F). This observation appears, however, implausible 777 considering that NW-SE oriented coastlines may be exposed more directly to the erosive effect of storm waves 778 associated with winds approaching from the south. We interpret the orientation of the coastline therefore to be of 779 secondary importance at regional scale for the formation of marine terraces compared to other parameters, such as 780 wave height, tidal range, or rainfall. 781

Conclusions 796
We measured 1,953 shoreline-angle elevations as proxies for paleo-sea levels of the MIS-5e and 5c terraces along 797 ~5,000 km of the WSAC between Ecuador and Southern Chile. Our measurements are based on a systematic 798 methodology and the resulting data have been standardized within the framework of the WALIS database. Our 799 mapping was tied using referencing points based on previously published terrace-elevation estimates and age 800 constraints that are summarized in the compilation of Pedoja et al. (2011). The limitations of this database are 801 associated with the temporal accuracy and spatial distribution of the referencing points, which we attempt to consider 802 by providing a quality-rating value to each measurement. The marine terrace elevations display a median value of The pattern of terrace elevations displays short-, medium-and long-wavelength structures controlled by a combination 809 of various mechanisms. Long-wavelength structures may be controlled by deep-seated processes at the plate interface, 810 such as the subduction of major bathymetric anomalies (e.g. Manta Peninsula and San Juan de Marcona region). In 811 contrast, short-and medium-wavelength deformation patterns may be controlled by crustal faults rooted within the 812 upper plate (e.g., between Mejillones and Valparaíso). 813 Latitudinal climate characteristics along the WSAC may influence the generation and preservation of marine terraces. 814 An increase in wave height and tidal range generally results in enhanced erosion and morphologically well-expressed, 815 sharply defined marine terraces, which correlates with the southward increase in the number of our marine terrace 816 measurements. Conversely, river incision and lateral scouring in areas with high precipitation may degrade marine 817 terraces, thus decreasing the number of potential marine terrace measurements, such as observed south of Valparaíso.