Contemporary Salt-Marsh Foraminifera from Southern California and Implications for Reconstructing Late Holocene Sea-Level Changes

Studies of sediment successions from salt marshes provide opportunities for quantitative reconstructions of late Holocene relative sea- and land-level changes around the world (Guilbault et al., 1996; Gehrels et al., 2004; Sawai et al., 2004; Kemp et al., 2011). Quantitative sea-level studies rely on modern faunal or floral zonation with respect to elevation over a range generally covering the intertidal zone and occasionally extending into subtidal environments (e.g., Horton et al., 1999; Gehrels et al., 2001; Woodroffe, 2009; Avnaim-Katav et al., 2016). Once the relationship between the relative abundance of different species of a particular microfossil group and elevation is established, it is used to develop predictive transfer functions capable of inferring the past elevation relative to the tidal frame from fossil records (Barlow et al., 2013; Kemp & Telford, 2015). Within salt-marsh sedimentary archives, benthic foraminifera are among the most commonly used microorganisms for reconstructing past sea-level changes due to their vertical zonation in the modern intertidal zone (e.g., Scott & Medioli, 1980; Scott et al., 1984; Gehrels, 1994; Horton et al., 1999; Kemp et al., 2009; Horton & Edwards, 2005). They work particularly well in micro- to mesotidal settings (Barlow et al., 2013; Kemp & Telford, 2015; Williams et al., 2021). With sufficient sampling resolution, transfer functions can provide a near-continuous record of sea-level change with precision of ∼0.1–0.3 m (Williams et al., 2021).

Microfossil transfer functions are widely applied to reconstruct the direction and magnitude of abrupt coseismic changes in land level, which are experienced along coastlines as sudden changes in relative sea level (Hocking et al., 2017; Shennan et al., 2016; Brader et al., 2021). The Newport–Inglewood–Rose Canyon fault zone is a large and active fault system that extends along ∼70 km of the southern California coast (Sahakian et al., 2017). Substantial vertical deformation may not be expected along this strike-slip fault system; nevertheless, intense shaking may still result in marsh submergence and abrupt relative sea-level rise due to lateral spreading and sediment consolidation (Aydan et al., 2008). Consequently, in areas subject to these processes, relative sea-level rise results in an increase in accommodation space, leading to distinct and observable changes in paleoecological records (Darienzo & Peterson, 1995; Nelson et al., 1996; Shennan et al., 2016). Such submergence of coastal salt marshes might be evident in the abrupt transition from a peaty soil to intertidal mud deposition, as shown in numerous studies exploring earthquakes along the Cascadia Subduction Zone (e.g., Darienzo & Peterson, 1995; Nelson et al., 1996; Shennan et al., 1996; Atwater & Hemphill-Haley, 1997).

In Southern California, Avnaim-Katav et al. (2017) developed a transfer function using a modern foraminiferal training set from two salt marshes. However, the range of elevations covered by this training set was limited, and no comparison with fossil foraminiferal assemblages was undertaken. Consequently, this study first aims to create a robust and extensive regional-scale modern training set from which transfer functions can be developed. To achieve this we describe the distribution of contemporary foraminiferal assemblages with regards to elevation in two southern Californian salt marshes, Kendall-Frost Mission Bay Marsh Reserve and Carpinteria Slough, before merging them with the data of Avnaim-Katav et al. (2017). Secondly, we seek to test the applicability of the resulting transfer functions by investigating fossil foraminiferal assemblages from three cores from the Kendall-Frost Mission Bay Marsh Reserve near San Diego. We use quantitative reconstructions of changes in relative sea level to discuss initial interpretations of the sea-level, tectonic, and anthropogenic history of the site and identify directions for future palaeoseismic and relative sea-level research along this populous coastline.

Studies on foraminiferal ecology in Mission Bay date back to the 1960s and mid-late 1970s. Phleger & Bradshaw (1966) discussed the low species richness within intertidal environments compared to subtidal environments, associated with rapid changes in their abiotic parameters over a tidal cycle. Scott (1976) studied the recent paleoecology of this and other southern Californian salt marshes. Scott et al. (2011) integrated these earlier results with borehole sediments aimed to reconstruct Holocene paleoenvironmental changes; however, cores were collected close to an old river channel that may have biased the stratigraphic record.

Pristine, unaltered salt marshes suitable for sea-level reconstruction are lacking along the southern Californian coastline. Human activity has greatly reduced the extent of salt-marsh environments and caused substantial geomorphic changes, as detailed above. Nevertheless, through the application of palaeoenvironmental reconstruction approaches, it may be possible to disentangle the anthropogenic impacts and other processes including long-term sea-level change and abrupt changes associated with earthquakes. Such abrupt changes associated with coseismic deformation or intense shaking are anticipated given the close proximity of the site to the Newport–Inglewood–Rose Canyon fault zone, a large and active strike-slip fault system (Fig. 1). The fault zone is part of the broad, distributed shear zone comprising the Pacific-North American plate boundary. Most long-term slip in Southern California occurs on the San Andreas, San Jacinto, and Elsinore faults but 12–15% of the ∼50 mm/yr of plate boundary shear is attributed to the coastal and offshore faults, including the Newport–Inglewood–Rose Canyon fault zone (Sahakian et al., 2017).

Carpinteria Slough (34°24.0’N, 119°31.5’W) lies approximately 15 km east of Santa Barbara and 150 km northwest of the Seal Beach site investigated by Avanim-Katav et al. (2017) (Fig. 1). The 230-ha marsh consists of three basins; the two eastern basins are separated from the western basin by an artificial road. Channels within the eastern half of the marsh are channelized and dredged, resulting in a highly altered tidal flushing of the marsh plain (Sadro et al., 2007). The western basin contains a much more complex array of tidal channels and creeks, which are largely unaltered (Sadro et al., 2007); therefore, we focused our surface sampling on this area. Tidal input and drainage primarily occurs through a tidal inlet, which sits at the southern margin of the marsh (Sadro et al., 2007). The marsh is separated from the open Santa Barbara Channel by a spit that has been breached during large multi-century storms (Reynolds et al., 2018).

Mission Bay and Carpinteria Slough share similar climatic and oceanographic settings, representative of Southern California. Tides along the open coast are semidiurnal and have a mean range of ∼1.2 m. We discuss site-specific tidal ranges in the ‘METHODS’ section.

The marshes demonstrate a vegetation zonation typical of the marshes of Southern California (Zedler, 1977; Zedler et al., 1986). The lowest vegetated zone is characterized by Spartina alterniflora. The mid-marsh zone is commonly co-dominated by Sarcocornia pacifica, Batis maritima, and Jaumea carnosa. Distichlis spicata, Frankenia grandifolia, and Limonium californicum occur in the mid-to-high marsh areas. The boundary between mid and high marsh vegetation is less well-defined, and plants from these zones also inhabit the marsh-upland transition area. Arthrocnemum subterminale and Monanthochloe littoralis prefer the highest marsh elevations. The marsh-upland transition is marked by the presence of shrub-type plants, such as Artemisia californica, Rhus lauriana, and Baccharis pilularis.

The climate of Southern California Is Mediterranean (Xeric) and characterized by hot, dry summers and cool to warm, wet winters. The average annual high and low temperatures are 21°C and 14.1°C, and the annual precipitation average is 26.4 cm (usclimatedata.com). Drought and extreme flooding are common and pose substantial climatic impacts on marsh vegetation (Zedler et al., 1986; Zedler, 2010; Reynolds et al., 2018). During the dry season (March to September) marsh soils are characteristically hypersaline because most of the soil moisture originates from tidal inundation and because evaporation typically exceeds precipitation (Zedler, 1982).

To assess the distribution of tidal-marsh foraminifera we collected 11 surface samples from Mission Bay and 29 from Carpinteria Slough. At both sites, we established a linear sampling transect (Fig. 1; Appendix 1). At Carpinteria Slough, this was supplemented by an additional broad swath of samples (17 in total, prefaced CS_ES; Fig. 1). Sampling at Mission Bay incorporated low, mid, and high marsh vegetation zones, with samples from Carpinteria Slough also extending down into the upper part of the unvegetated tidal flat. Sampling along transects (e.g., within one site/marsh) may lead to spatial autocorrelation and thus might negatively affect transfer function model performance (Legendre & Fortin, 1989; Telford & Birks, 2009). The issue of spatial autocorrelation is reduced by including samples from different salt marshes and combining the results into one modern training set. To this end, we combine the 40 samples reported in this paper with 51 modern surface samples reported by Avnaim-Katav et al. (2017) from marshes in the Seal Beach and Tijuana estuaries (Fig. 1).

At each surface sampling location, we sampled a standardized volume of 10 cm³ from the topmost centimeter of salt-marsh sediment (10 cm² by 1 cm thick) for foraminiferal analysis. This sampling strategy has been widely used in many marshes studied around the world, although it does not capture the presence of infaunal species (Scott & Medioli, 1980; Patterson et al., 2004). Sample preparation followed Schönfeld et al. (2012) and Avnaim-Katav et al. (2017). Samples were stained with rose Bengal solution (2 g rose Bengal/l 95% ethanol) at the time of sampling. Rose Bengal confirms the presence of cytoplasm and is widely used to distinguish between dead and assumed living foraminifera. As cytoplasm may potentially persist for years or decades after death (Bernhard et al., 2006), “live” counts may also include recently living foraminifera. Nevertheless, comparing rose Bengal with the vital stain CellTracker Green, Figueira et al. (2012) suggested no significant difference between the ability of the two staining techniques to discriminate between live and dead foraminifera in tidal marsh samples. Specimens were included in an “assumed living at the time of collection” category when multiple chambers were stained bright red (Horton & Edwards, 2006; Milker et al., 2015). The solution was buffered by calcium carbonate powder to avert dissolution of calcareous species. Samples were wet sieved through 500-μm and 63-μm sieves. The >500-μm fraction was analyzed for larger foraminifera before being discarded. The fraction between 63 and 500 μm was subdivided into eight equal aliquots using a wet splitter (Scott & Hermelin, 1993).

Tests were counted in water, which enabled easy detection of rose Bengal-stained foraminifera in the surface samples and prevented drying of the organic residue. Samples with more than 50 dead tests were used in numerical analyses, following Kemp et al. (2020). Two samples from Carpinteria Slough were excluded from the analysis due to low counts. Consequently, eighty-nine samples were included in the modern training set.

Taxonomic identifications follow the World Foraminifera Database (Hayward et al., 2022). Juvenile specimens of Trochamminids were lumped into a single group because they were difficult to identify to species level due to their small sizes but excluded from statistical analyses as they reflect a range of species that inhabit different elevations. All counts were expressed as a relative abundance (%). Scanning electron microscope photographs of key species were taken at the Department of Earth, Planetary and Space Sciences, UCLA.

The locations of the surface samples were determined with a Differential Global Positioning System (dGPS). Each location was post-processed with the Online Positioning User Service produced by the US National Oceanic and Atmospheric Administration (NOAA) to standardize and correct for spatial changes in mean sea level, MSL. The elevations, with a precision averaging 4 cm, were referenced to the North American vertical datum (NAVD88) computed using GEOID12B.

Following Avnaim-Katav et al. (2017), we use NOAA's vDatum tool (http://vdatum.noaa.gov/) to assess tidal datums [mean higher high water (MHHW), mean tide level (MTL), and mean lower low water (MLLW)] at Mission Bay salt marsh (Table 1 ). While vDatum provides datums for the open coastline, the enclosed nature of Mission Bay could result in modification of the tidal amplitude. Nevertheless, NOAA predictions for Quivira Basin, within Mission Bay and around 3 km south of the sampled marsh, suggest no tidal dampening occurs (NOAA, 2022).

Table 1.

Tidal datums for each study site including Mean Higher High Water (MHHW), Mean Tide Level (MTL), and Mean Lower Low Water (MLLW). Datums for Carpinteria are from Sadro et al. (2007); those from Tijuana are based on five years of tidal data from publicly available water level data (http://cdmo.baruch.sc.edu/dges/). Datums for Mission Bay and Seal Beach are generated using the National Oceanic and Atmospheric Administration (NOAA) vDatum tool.

View largeDownload slide

View Large

For Carpinteria Slough, we use datums derived by Sadro et al. (2007) from a vented pressure sensor deployed between August 2005 and April 2006 (Table 1). While Avnaim-Katav et al. (2017) used vDatum to derive tidal datums for Tijuana, we update their analysis using five years of data from the National Estuarine Research Reserve System (http://cdmo.baruch.sc.edu/dges/). We use water-level data from Oneonta Slough for the period from September 2016 to August 2021 to define MTL and MHHW (Table 1). This reanalysis results in a smaller tidal range than used in the previous publication.

We convert the elevation of each modern sample to a standardized water level index (SWLI), which allows us to combine samples from sites with different tidal ranges into a single training set (Gehrels, 1999; Barlow et al., 2013; Kemp & Telford, 2015). Wright et al. (2011) recommend using the highest occurrence of foraminifera as the upper reference level in SWLI calculations. However, we could not establish this datum because our highest samples still contained foraminifera. Consequently, we used MHHW as the upper reference level. Our index assigns MHHW at each site a SWLI value of 200 and MTL a value of 100.

We assume that the tidal range in Mission Bay has not changed over time. Natural and anthropogenic modification of Mission Bay means that this assumption may not be accurate; however, we do not have any data to support the use of alternative tidal ranges when converting SWLI values back into meters. The consequence of this is that rates of past changes may be overestimated or underestimated. In particular, for abrupt changes in relative sea level, the magnitude of the change may be miscalculated. Consequently, all such estimates should be viewed as initial approximations and potentially subject to later recalculation if new evidence for tidal range changes arises.

Dead surface foraminiferal assemblages, rather than total or live assemblages, were statistically analyzed to minimize the influence of seasonal fluctuations (Culver & Horton, 2005; Milker et al., 2015). While the living assemblage represents a single point in time and may be influenced by seasonal blooms, dead assemblages provide time-averaged information about the assemblages, and thus most accurately reflect the subsurface assemblages. The following statistical methods were applied to the combined foraminiferal and elevation data from four salt marshes: Mission Bay, Carpinteria Slough, Tijuana, and Seal Beach.

In order to classify the distribution of groups and subgroups in the foraminiferal samples into homogeneous faunal zones (clusters) we used a Q-mode cluster analysis in PRIMER version 6 software (Clarke & Gorley, 2006) following steps described in Avnaim-Katav et al. (2017). Rare species (<1 % maximum relative abundance) were excluded from the analysis.

A detrended canonical correspondence analysis (DCCA; Ter Braak, 1986) was carried out to determine the type of response, unimodal or linear, displayed by the species to the elevation gradient. With gradient lengths (‘species turnover’) of >2 standard deviation units, our DCCA indicated a unimodal species response and therefore canonical correspondence analysis (CCA) was applied to quantify the relationship between the distributions of benthic foraminifera and elevation. Both DCCA and CCA were applied using Canoco, version 4.55 software (ter Braak & Šmilauer, 2002; Lepš & Šmilauer, 2003) following the steps detailed in Avnaim-Katav et al. (2017).

The relationship between elevation (SWLI) and the relative abundances of foraminiferal taxa was empirically modelled using transfer functions. We do not apply any species transformation or remove any outlying samples. Transfer functions were developed using Weighted Averaging (WA; ter Braak, 1987), Weighted Averaging Partial Least Squares (WAPLS; ter Braak & Juggins, 1993) and Locally Weighted Weighted Averaging (LWWA; Juggins & Birks, 2012) in C2 version 1.7.4 (Juggins, 2011). For LWWA we include 30 samples in the “local” training set and use the chord-squared distance metric. Transfer function performance was evaluated based on widely applied criteria (Birks, 1998; Juggins & Birks, 2012; Kemp & Telford, 2015). The model with the highest bootstrapped r² (1000 cycles) and the lowest root mean square error of prediction (RMSEP) value was chosen. For WAPLS, to avoid overfitting the data, improvements in RMSEP of less than 5% between successive model components were deemed insignificant (ter Braak & Juggins, 1993). This decision path was only applied to the first three components to limit complexity of the statistical analysis (Wright et al., 2011; Barlow et al., 2013).

We applied the foraminifera-based transfer function to fossil foraminiferal assemblages from three cores from Mission Bay (see the following section). We use the modern analogue technique (MAT) to quantify the resemblance between each fossil sample and the modern training set (Birks, 1998; Kemp & Telford, 2015). We selected the squared chord distance (Overpeck et al., 1985) for the calculation of minimum dissimilarity coefficient (MinDC). The 20th percentile of dissimilarity values between all possible pairings of modern samples was selected as the threshold (Watcham et al., 2013). Samples with dissimilarity coefficients lower than the 20th percentile were defined as having close analogues and samples with coefficients greater than the 20th percentile as having no close analogues.

To provide an initial test of the applicability of the modern training set for reconstructing depositional elevations from fossil salt-marsh sequences, we investigated the stratigraphy of the Mission Bay salt marsh. The subsurface sediment was described from 10 hand-driven cores collected in the northern and western parts of the marsh (Fig. 1; Appendix 2). These cores were collected in overlapping, 50-cm-long sections using a Russian corer to avoid compaction. Selected core sections were wrapped, labelled, and kept alongside all surface samples in a refrigerator at 4°C prior to laboratory analyses. The stratigraphy was documented along two transects: an east to west transect and a north to south transect (Fig. 1D). The latter transect is oriented in the same direction as cores described by Scott et al. (2011). The stratigraphy is based on the characterization of the sedimentary facies described in all cores, and the lateral stratigraphic relations between them. Each sedimentary facies was described based on its observable physical features, such as sediment type, color, texture (grain-size, sorting), and fossils.

Three cores from the eastern part of the marsh were selected for analysis. Their selection was based on a reconnaissance survey, analysis of aerial photography, and a historical map from 1852 that indicated minimal human modification during the 20th century and that the sites were far from river channels and/or tidal flats (Figs. 1D–E). Core MB17-05 was selected for foraminiferal analysis and dating on the basis of its position in the high marsh zone, and thus it includes the thickest and most continuous sequence of salt-marsh sediment accumulation overlying incompressible basement rock. Cores MB17-08 and MB17-07 were chosen for comparison with MB17-05, aiming to replicate the results from core MB17-05. Sampling for fossil foraminifera followed the approach applied to the surface samples.

We developed a chronology for cores MB17-05 and MB17-08 based on ¹⁴C dating, ²¹⁰Pb analysis and ¹³⁷Cs age markers. For ¹⁴C dating, we sampled identifiable aboveground material from terrestrial plants to avoid erroneously young dates from root matter. Samples were washed in deionized water, dried at 70°C for 24 hours, and transported to the University of California, Irvine Keck Radiocarbon Laboratory, where they were cleaned, combusted, and graphitized before undergoing accelerator mass spectroscopy (AMS). Values of ¹³⁷Cs and ²¹⁰Pb were measured using an Ortec germanium crystal well detector at the PEARL Laboratory at Queen's University, Kingston, Ontario using methods from Schelske et al. (1994). A total of 18 samples were tested for gamma activity in core MB17-05, the thickest marsh record, and seven samples were tested along the core length of MB17-08 to determine the basic profile shape, to be compared with MB17-05. We developed Bayesian age-depth models using the rplum package (Blaauw et al., 2022) within the R environment (R Core Team, 2013), incorporating the ¹³⁷Cs, ²¹⁰Pb, and ¹⁴C results. This approach alleviates the need to remodel outputs from traditional ²¹⁰Pb depositional models (Aquino-López et al., 2018, 2020). Radiocarbon dates were calibrated within rplum using the IntCal20 calibration curve (Reimer et al., 2020). We included boundaries at key stratigraphic transitions to allow greater model flexibility and adjusted depths to account for any abruptly deposited units.

Foraminiferal assemblages in cores MB17-07 and MB17-08 are also dominated by T. inflata and E. macrescens, with A. mexicana encountered below 50 cm core depth (Figs. 7B, C). As in the equivalent layer of facies #4 in core MB17-05, M. fusca increases in the upper light grey organic sandy-clayey silt in cores MB17-07 and MB17-08. The lower layer of this facies in core MB17-08, which is absent in MB17-05, displays an abrupt increase in M. fusca coincident with the stratigraphic boundary. Although sampled at lower resolution, this abrupt change is replicated in core MB17-07, where M. fusca is absent in the underlying sediments but exceeds 50% of the assemblage in the light grey silt unit.

Activity of ¹³⁷Cs is low through most of core MB17-05 with a small peak between 28–36 cm depth. The rplum model indicates that this peak is an outlier and significantly younger than the ages inferred from the ²¹⁰Pb and ¹⁴C results (Fig. 8A). Low ¹³⁷Cs fallout means that records from Southern California can be unreliable (Drexler et al., 2018); however, the occurrence of ¹³⁷Cs at deeper than expected depths may also point towards post-depositional mobility in the sediment profile (Foster et al., 2006). Core MB17-08 presents a clearer peak in ¹³⁷Cs activity at 15 cm depth (Fig. 8B), and the age-depth model indicates that this is consistent with the other chronological data from this core.

The tidal flat and low marshes of the study sites are characterized by the presence of the agglutinants M. fusca and S. moniliformis and several calcareous species. These agglutinants are known as lower-elevation species along, for example, the North American Pacific and Atlantic coasts (e.g., Edwards et al., 2004; Nelson et al., 2008; Milker et al., 2015; Avnaim-Katav et al., 2017). In the middle marsh and the high marsh, the dominant species include T. inflata and E. macrescens. Their distribution is similar to other studies. For example, in Australia and other areas in North America, E. macrescens and T. inflata have been reported from middle marsh environments (Guilbault et al., 1996; Nelson et al., 2008; Hawkes et al., 2010; Engelhart et al., 2013, Milker et al., 2015) and the highest marsh positions (e.g., Patterson, 1990; Gehrels & van de Plassche, 1999; Hippensteel et al., 2002; Woodroffe & Horton, 2005). Some of the highest elevation samples in this study included T. irregularis and M. petila. These species have frequently been identified as the dominant species in the high and highest positions within North American Pacific salt marshes (Hawkes et al., 2010; Engelhart et al., 2013; Milker et al., 2015), as well as in New Zealand (Southall et al., 2006) and Tasmania (Callard et al., 2011). Likewise, M. petila has also been reported in the middle and high marshes from Oregon (Engelhart et al., 2013).

The zonation in modern foraminiferal assemblages from the four salt marshes confirms their potential for being used as accurate and precise indicators of depositional elevations. Clusters of samples defined by Q-mode cluster analysis overlap but can be related to different marsh elevation zones (Fig. 4). While elevation explains 16% of the variance in the regional training set, this result is similar to other studies from Canada, Oregon, USA and the UK (e.g., Barnett et al., 2016; Hawkes et al., 2010; Horton & Edwards, 2006), and not unexpected given that other variables, such as salinity and grain size, also influence foraminiferal distributions (Horton et al., 1999; Horton & Edwards, 2006; Avnaim-Katav et al., 2017).

The RMSEP, 22 SWLI units, is equivalent to 0.18 m at Mission Bay and is 16% of the sampled elevation range. These values accord well with the ranges discussed by Callard et al. (2011), Barlow et al. (2013), and Williams et al. (2021).

Reliable reconstructions of paleo-marsh elevation and consequently relative sea-level or land-level change depend on the presence of similar assemblages in the contemporary and fossil data sets and that the species’ responses to the environment haven't changed during the course of time. Our MAT results show that the modern training set provides 69 out of the 101 fossil samples in the three cores from Mission Bay with close analogues (Figs. 7). The largest number of poor analogues occur in the lower parts of the high marsh core, MB17-05, at the eastern end of the transect (23 poor analogues out of 56 samples; Fig. 7A). This is largely due to the presence of Arenoparrella mexicana, which is absent in the modern data set. This species is known from mangrove environments living in the subtidal zone, for example in southern Brazil (Barbosa et al., 2005) or in the upper intertidal zone in Australia (Berkeley et al., 2008). It is also found in the Great Marshes of Massachusetts; however, there it does not demonstrate vertical zonation (de Rijk & Troelstra, 1997). In a Georgia salt marsh, it is more abundant as an infaunal species in the high-marsh subsurface than in surface samples where it is generally absent (Goldstein & Harben, 1993). In the Morbihan Gulf marshes along the Atlantic coast of SW Europe, this species was recorded mostly in stations slightly below MHHW (Leorri et al., 2010). The absence of consensus in the literature about the distribution of this species and the potential for infaunality mean samples containing this species should be treated cautiously when estimating paleo-marsh elevations. Future research should seek to resolve the distribution of A. mexicana in southern California.

We use rplum to create age-depth models which integrate results from ¹³⁷Cs, ²¹⁰Pb, and ¹⁴C methods. As suitable radiocarbon samples closely bracketing E1, E2, and E3 were not encountered and the depositional environment of California coastal marshes can present challenges when using ¹³⁷Cs or ²¹⁰Pb (Drexler et al., 2018), the ages presented here for the events remain somewhat uncertain. In particular, the timing of E2 in core MB17-05 (estimated by the age-depth model as 1805–1894 CE) is poorly constrained as it lies below the limit of significant ²¹⁰Pb activity and more than 30 cm above the nearest radiocarbon date. If this sedimentary transition is indeed synchronous between cores, the age derived from core MB17-08 (1885–1934 CE) would be preferred as it is better constrained by the available chronohorizons. A more conservative interpretation would place this transition before ∼1900 CE—the time at which excess ²¹⁰Pb can be detected (Jeter, 2000)—but after 1530–1794 CE.

While our models do make allowances for abruptly deposited units, particularly E1 as discussed in the following section, the lack of ages closely bracketing the events prevents us from making inferences about the duration of any hiatuses in sediment deposition or the occurrence of erosion. Future sampling could focus on obtaining dates from immediately above and below each boundary; this would allow alternative approaches such as Sequence modelling to refine the age estimates (cf. Lienkaemper & Bronk-Ramsey, 2009; Shennan et al., 2014).

We use the transfer function outlined above (section ‘Transfer function development’) to calibrate the fossil assemblages in the three cores (Fig. 7). Throughout the salt-marsh sequence, reconstructed paleo-marsh surface elevations range from approximately 130 to 270 SWLI units, equivalent to ∼1–2 m above MTL, assuming that the tidal range has not changed over time. While these marsh surface elevations could be combined with the age-depth model described in section ‘Core chronology’ to realize a relative sea-level reconstruction, such a step is complicated at present by our incomplete knowledge of geomorphic changes at the site. Such changes, including those linked to rerouting of the San Diego River, sediment infilling, and dredging may have resulted in substantial changes to the tidal range over time and therefore altered the conversion of SWLI values into meters. Consequently, rather than focusing on the long-term changes in relative sea level, we concentrate on the origin of the intercalated layers within sedimentary facies #2.

Three notable changes in stratigraphy occur within the salt-marsh sediments of sedimentary facies #2. The oldest transition is recorded in mid and low marsh cores, including MB17-07 and MB17-08, as a change in the color of the organic sandy-clayey silt, accompanied by an abrupt increase in the relative abundance of the low marsh species M. fusca and a decline in the mid to high marsh E. macrescens and T. inflata (Figs. 7B, C). Assuming no simultaneous change in tidal range, reconstructed palaeomarsh surface elevations indicate an abrupt decrease in paleomarsh surface elevation of 0.71±0.26 m in core MB17-08 and 0.41±0.24 m in core MB17-07. It should be noted that the reconstruction for core MB17-08 is reliant on a sample that lacks a close modern analogue due to the contribution of A. mexicana (discussed in section ‘Comparison of modern and fossil assemblages’). The age model for MB17-08 indicates this transition, which we refer to as event 3 or E3, occurred at 215–284 cal yr BP (1666–1735 CE; Fig. 8).

A second change in marsh stratigraphy is recorded in all three sampled cores, again primarily as a change in the color of the organic sandy-clayey silt. While cores MB17-05 and MB17-08 both see an increase in M. fusca across the contact, the increase in this low marsh species is gradual, starting 10 cm below the change in sediment color in MB17-05. A corresponding increase in M. fusca in core MB17-07 is also evident, but the abrupt or gradual nature of the change is less clear due to the lower sampling resolution. Reconstructed paleomarsh surface elevations show a gradual decrease ranging between ∼0.2 and 0.5 m in the three cores analyzed. The age models suggest that the stratigraphic transition, E2, occurred at 56–145 cal yr BP (1805–1894 CE) in MB17-05 and 16–65 cal yr BP (1885–1934 CE) in MB17-08 (Fig. 8).

A third environmental change, E1, is recognized as an organic–minerogenic contact occurring only at the eastern edge of the marsh in core MB17-05 (Fig. 7A). The silty-clayey sand deposits of sedimentary facies # 3 lack lateral continuity between cores and are devoid of foraminifera. Under and overlying sediments contain similar assemblages dominated by the mid to high marsh foraminifera E. macrescens and T. inflata, with no difference in reconstructed paleomarsh surface elevation. The age model indicates the sand was deposited between -10 and 45 cal yr BP (1905–1960 CE; Fig. 8).

Coseismic deformation on the Newport–Inglewood–Rose Canyon fault zone is unlikely to produce any significant vertical deformation in the study area due to the strike-slip fault motion (Lindvall & Rockwell, 1995; Rockwell, 2010). Nevertheless, earthquake shaking-induced lateral spreading could result in abrupt changes in paleomarsh surface elevation. Such an occurrence would be expected to share characteristics more usually associated with upper plate deformation in subduction zone settings, including lateral continuity of stratigraphic evidence between cores, evidence for a sudden change in marsh elevation and synchroneity with other evidence for earthquake occurrence (Nelson et al., 1996; Shennan et al., 2016). As E3 can be traced across multiple cores and displays foraminiferal evidence for an abrupt, decimeter-scale decrease in marsh elevation, we hypothesize that the stratigraphy might record the secondary effects of earthquake shaking. Paleoseismic trenching across the Rose Canyon fault in the Old Town area of San Diego documented four large and two smaller late Holocene surface-rupturing events, with the most recent earthquake occurring between 1704 CE and the mid-1700s (Grant & Rockwell, 2002; Singleton et al., 2019). The timing of this earthquake correlates well with our preliminary age assessment for E3 (1666–1735 CE). We suggest shaking-induced lateral spreading could have resulted in an abrupt decimeter-scale decrease in marsh surface elevation, and a sudden consequent change in foraminiferal assemblages in the Mission Bay marsh. A similar analogue for such a mechanism occurred during the 1999 İzmit earthquake, Eastern Turkey, where liquefaction-induced lateral spreading resulted in submergence of coastal areas (Aydan et al., 2008). While the event predates the start of substantial anthropogenic impacts in Mission Bay, further work is needed to rule out the possibility of other causal processes, such as non-seismic geomorphic changes or anthropogenic impacts. Additional mapping of the stratigraphic change and further quantification of the amount and abruptness of the possible elevation change from additional cores would also help to support or refute the hypothesized process.

The foraminiferal changes associated with E2 and E1 are not consistent with the mechanism proposed for E3. Event 2, tentatively dated to the 19^th to early 20^th century, lacks evidence for an abrupt change in marsh elevation. Instead, the gradual change observed in assemblage composition, including the increased abundance of M. fusca that dominates lower elevations may reflect anthropogenic changes to the geomorphology and tidal range in Mission Bay, in particular following the rerouting of the San Diego River's course into the bay in 1850 (Marcus, 1989). As E1 lacks widespread lateral continuity and evidence for marsh surface elevation change, fluvial flooding provides one plausible formation mechanism. The age range (1905–1960 CE) coincides with the 1938 CE floods, which resulted from a series of tropical storms (the Lash of Saint Francis) that were the largest floods in the history of southern California and covered coastal areas of Los Angeles, Orange, and Riverside Counties with >1-m thick deposits of silty sand (Poland & Piper, 1956). Further mapping and sedimentological analyses could help to confirm the origin of this layer.

To investigate the control of elevation, a surrogate for the frequency and duration of tidal inundation, on the distribution of salt-marsh foraminifera in Southern California, we quantified faunal assemblages in surface samples along transects from marshes at Mission Bay and Carpinteria Slough. We combined these samples with existing datasets from Seal Beach and Tijuana to form an 89-sample regional modern training set. Unconstrained cluster analysis divided the foraminiferal assemblages into four elevation-dependent faunal zones. Elevation exerts a major control over species distributions, explaining 16% of the variance in the dataset, and species are unimodally distributed along the elevation gradient. Consequently, we developed transfer functions to explore the potential for the use of the training set for quantifying past changes in marsh surface elevation. Our chosen model displays a strong relationship between observed and predicted marsh elevations in the modern data set. With prediction uncertainties equivalent to ±0.18 m at Mission Bay, the transfer function provides precise reconstructions of paleo-marsh surface elevation from appropriate fossil foraminiferal assemblages.

We perform an initial test of the applicability of the training set and associated transfer function by investigating fossil foraminifera preserved in sediment successions from the Mission Bay site. Here, hand-driven cores map the marsh stratigraphy and have the potential to provide evidence for late Holocene paleoenvironmental changes. The modern training set provides close analogues for the majority of fossil foraminiferal samples; however, many fossil samples contain Arenoparrella mexicana, a species not found in the modern samples. This disparity may reflect the infaunal living position of the species, which is not captured by the modern sampling approach. The Mission Bay site is not ideally suited to long-term relative sea-level reconstruction due to substantial anthropogenic modification and potential but unknown changes in tidal range over time. Instead, we limit our focus to three notable stratigraphic changes within the salt-marsh sediments. The abrupt and widespread change in stratigraphy, increase in low marsh taxa and quantitative reconstructions indicating decimeter-scale submergence may point towards the oldest change being associated with lateral spreading of the water-saturated sediments due to strong ground shaking from a local earthquake. Bayesian age-depth modeling indicates that this event might be contemporaneous with a moderate to large-magnitude earthquake that occurred on the Rose Canyon fault in the early to mid-18^th century CE; however, further dating is required to more precisely constrain the age of the sedimentary evidence. Two younger stratigraphic changes are not consistent with this mechanism. A gradual increase in low marsh taxa in the mid 19^th to early 20^th century may relate to rerouting of the San Diego River into Mission Bay and resulting changes in sediment infilling and tidal range, while a fauna-barren sand deposit encountered towards the landward edge of the marsh may reflect fluvial flooding.

The results presented here indicate the potential for our new foraminifera-based transfer function to support studies of the palaeoenvironmental evolution of the Southern Californian coastline. While the interpretation of the Mission Bay sequence remains preliminary due to the complexity of the site and the potential for unrecognized changes in tidal range, it appears likely that coastal sediments in this region may preserve evidence for the environmental effects of large earthquakes. Recognizing geological evidence for these events is potentially of considerable value for informing future hazard assessment and mitigation strategies in the densely populated Southern California. Future coastal studies should therefore seek to provide improved constraints on the magnitude and frequency of earthquakes in this region.

This study was supported by the US Department of Interior establishing grant #Y561461:03 for the Southwest Climate Science and by the Southern California Earthquake Center (Contribution No. 11012). SCEC is funded by NSF Cooperative Agreement EAR-1600087 & USGS Cooperative Agreement G17AC00047. We gratefully acknowledge the Kendall-Frost Mission Bay Marsh Reserve and Isabelle Kay the Manager and Academic Coordinator of the Natural Reserve System UC San Diego for site access, helpful suggestions. Thanks are due to Matthew E. Kirby (Cal-State Fullerton) for the granulometric analysis and Pamela Buzas-Stephens and Martin A. Buzas for help with foraminiferal analysis in Carpinteria Slough. Matt Zebrowski of the Cartography Laboratory of the Department of Geography and Noah Bodzin of the Nanoelectronics Research Facility Department at UCLA (University of California Los Angeles), are thanked for the graphics and the SEM, respectively. Appendices 3–5 can be found linked to the online version of this article.