FATES (Fate of Terrestrial/Nonterrestrial Sediments) Program was conducted to understand processes and responses of substances from the land to the marine sink. A suite of 13 short box cores were collected between 2004 and 2006 from the continental slope off SW Taiwan, northern SCS (Table 1; (Huh et al., 2009; Liu et al., 2009). Locations of sediment cores are shown on the bathymetric map (Figure 1), and information regarding geographic coordinates, water depths, and core lengths are listed in Table 1. Sediments were sampled at 5-cm intervals throughout sediment cores for foraminiferal isotope analyses. Planktonic shell sizes used for isotopic measurements were constrained by the sieving mash size of 300–355 μm for Globigerinoides sacculifer (so-called Trilobatus sacculifer) to minimize ontogenetic effects. Benthic foraminiferal shells of Uvigerina sp. were picked from a fraction greater than 150 μm. Stable isotopic analyses were done on groups of 10 specimens or less for each sample. The picked foraminiferal specimens were cleaned thoroughly in an ultrasonic bath with methanol to remove adhering fine particles, followed by soaking in sodium hypochlorite (NaOCl, 5%) at room temperature for more than 24 h to further remove any fine organic particles. Cleaning with deionized distilled water followed, and samples were then oven dried at 50°C. Stable isotope analyses for specimens were measured at the Stable Isotope Laboratory, National Taiwan University, Taiwan, following standard procedures with a precision better than 0.07‰ for δ ¹⁸O and 0.04‰ for δ ¹³C.

Samples used for the radionuclide isotope analysis were sliced into sections with 2-cm intervals from the top of the sediment core. Each section was sealed in a plastic bag and stored in the refrigerator at 4°C before being freeze dried. After the sample was dried, the water content in the sediment sample was determined. Dried samples were then transferred to plastic jars (inside diameter is 8.5 cm; height is 7.5 cm) for nondestructive gamma spectrometric assay of radionuclides. Analyzed radionuclides include ²¹⁰Pb, ²¹⁴Pb, and ¹³⁷Cs, which were used as sediment chronometers (Huh et al., 2009) and were calculated according to a salt-free dry weight. The digital gamma-ray spectrometer was connected to HPGe detectors for counting radionuclides simultaneously based on photon peaks centering at 46.52 (²¹⁰Pb), 351.99 (²¹⁴Pb), and 661.62 (¹³⁷Cs) keV, respectively. Afterward, the counting results were analyzed with GammaVision 32 software (Su and Huh, 2002; Huh et al., 2006; Huh et al., 2009).

²¹⁴Pb is the precursor of ²¹⁰Pb and used as the index of supported ²¹⁰Pb (²¹⁰Pb_sup). An excess of ²¹⁰Pb (²¹⁰Pb_ex) can be obtained by subtracting the ²¹⁴Pb active concentration from the recorded ²¹⁰Pb (²¹⁰Pb_ex = ²¹⁰Pb_mea − ²¹⁰Pb_sup; (Huh et al., 2006; Huh et al., 2009). ¹³⁷Cs could be measured simultaneously with ²¹⁰Pb by nondestructive gamma spectrometry, but it took a long time to obtain ¹³⁷Cs data due to the lower activity. Therefore, only a portion of the cores were analyzed for ¹³⁷Cs to constrain the ²¹⁰Pb chronology. More details of the gamma spectrometry analysis are described in Huh et al. (2009).

Sedimentation rates derived from fallout radionuclides (²¹⁰Pb and ¹³⁷Cs) indicate constant hemipelagic accumulation, which implies that collected cores are suitable for reconstructing the recent paleoenvironment (Huh et al., 2009). The two out of 92 sediment cores (789-L1 and 791-K38) were selected as representatives in our study because of ample planktonic and benthic foraminiferal shells and the well age dating of box cores.

Ages of core samples at different depths were estimated by the sedimentation rate of each core. The linear regression of each core between δ ¹³C values of foraminifera shells and ages of samples was performed by the least square method using FITLM function provided by MATLAB R2020 software (Street et al., 1988). The two datasets, age later than 1900s and later than 1960s, were screened for regression analyses to compare with published records. Results of regression are listed in Tables 2 and 3 (e.g., slope, intercept, r squared, p value, and number of samples).

Coral skeletal records from Dongsha (Ren et al., 2017) and Liuqiu were also included to compare the anthropogenic effects on the two reef sites (Figure 1). The coral skeletal core collected around Liuqiu (120.36°E, 22.35°N) was drilled from a living Porites sp. colony at a depth of about 10 m on June 29, 2017. The core was cut into two slabs, then scanned by x-rays to identify its maximum growth axis, and then subsampled with an automated three-axis saw machine. Skeletal pieces (11 mm × 1 mm × 2.5 mm) with about monthly resolution were cut along the maximum growth axis and crushed into powder. About 5% of the powder is used for δ ¹³C and δ ¹⁸O analyses with isotope ratio mass spectrometer in the Stable Isotope Laboratory at the Department of Earth Science in National Normal University. The remaining powder is saved for other analysis. Combined with annual growth bands in x-ray images, skeletal δ ¹⁸O was compared with the extended reconstructed sea surface temperature (ERSSTv5) of the NOAA to establish the age model.

In general, ²¹⁰Pb in marine sediments is influenced by two factors: 1) the decay of ²²⁶Ra of marine sediments, ²¹⁰Pb_sup, and 2) the decay of atmospheric ²²²Rn, which deposits into the ocean and is preserved in marine sediments by the removal of particles from the water column, expressed as ²¹⁰Pb_ex (Hung and Chung, 1998). The depletion of ²¹⁰Pb_ex causes the ²¹⁰Pb activity to decrease exponentially or quasi-exponentially along the downcore until total ²¹⁰Pb activity in the sediment is equal to ²¹⁰Pb_sup. In short, the ²¹⁰Pb_ex profile represents the stable deposition if ²¹⁰Pb_ex decays exponentially with the depth and can be explained by the steady-state advection-decay model (Huh et al., 2009).

Assuming the amount of the sediment and ²¹⁰Pb flux at a given site are constant, and the mixing in the sediment is ignored. The downcore of ²¹⁰Pb_ex profile is, thus, invariable with time and described by: ln ⁡ C z = ln ⁡ C 0 − λ S Z (1) where C ₀ and C _z are ²¹⁰Pb_ex at the sediment–water interface and depth Z, respectively. λ is the decay constant (0.0311 year⁻¹) of ²¹⁰Pb, and S is the sedimentation rate (cm year⁻¹). Based on the regression analysis of ²¹⁰Pb_ex (Eq. 1), the linear sedimentation rate can be calculated (Holmes, 1998; Lewis et al., 2002; Huh et al., 2009).

Most of the ²¹⁰Pb_ex in the collected sediment cores decreased exponentially downcore with a few layers having low ²¹⁰Pb_ex values (Figure 2). The characteristics of ²¹⁰Pb_ex profiles were separated into Type I (sediment deposition in nonreworked settings; e.g., 789-L1), Type III (the existence of a physically and/or biological reworked surficial layer; e.g., 779-St11), and Type V (influenced by strong episodic events; e.g., 779-St9) according to the classification described by Xu et al. (2015). After excluding bio-interferences and event layers, the sedimentation rate and the preliminary age model of the 13 sediment cores were estimated.

¹³⁷Cs was introduced as the control point of the ²¹⁰Pb_ex-derived age model in collected sediment cores. The ¹³⁷Cs profile had the typical maximum radioisotope activity in the subsurface (the anthropogenic nuclide in 1963 A.D.) in which ¹³⁷Cs decreased gradually upward but sharply downward (Holmes, 1998; Huh et al., 2009; Figure 2). Since the ²¹⁰Pb-derived age model matches the control point indicated by ¹³⁷Cs, the quality of ²¹⁰Pb dating is reliable.

Additionally, the interference by the typhoon was considered as another control point of the age model in our study. For example, the intense rainfall caused flood deposits in the downstream shelf during Super Typhoon Haitang occurring on July 18–20, 2005 (Huh et al., 2009). This resulted in a low ²¹⁰Pb_ex layer occupying on the sediment core top collected afterward, especially for cores obtained in the same year of the typhoon event (e.g., 779-St9 and 779-St11 in Figure 2). The ²¹⁰Pb_ex profile implies the typhoon event did not interfere with the ²¹⁰Pb-derived age model severely in the long-term scale because the event layer (indicated by the minimum ²¹⁰Pb_ex value) was gradually buried by accumulating pelagic or hemipelagic sediments. Eventually, the event signal with low ²¹⁰Pb_ex was diluted and became weaker, though the signal was still observed in some of the collected sediment cores.

Oxygen and carbon isotopes of G. sacculifer in 13 box cores were plotted in Figure 3. Of the δ ¹⁸O data, 84% ranged between −2.5‰ and −2‰ for the last 150 years from 2010 (Figure 3A). The δ ¹³C records, however, fluctuated around 1.5‰ and started to decline after the 1900s followed by the rapid decline trend after the 1960s (Figure 3B). Both oxygen and carbon isotopic compositions from coretops were consistent with modern shells collected by sediment traps in this regime (Lin, 2014). To decipher historical isotopic changes, two out of 13 cores, 789-L1 and 791-K38, were selected as the representatives. The isotopes generated from planktonic foraminifera G. sacculifer and benthic foraminifera Uvigerina sp. are shown in Figures 4A–D. The planktonic isotopes of the other 11 cores are shown in Supplementary Figure S1.

The range and fluctuation pattern of planktonic δ ¹⁸O of two selected downcore records were very similar, particularly the broad δ ¹⁸O-enriched interval between 1950 and 1990 (Figure 4A). The deceasing of δ ¹⁸O occurring between 1950 and 1970 followed by a rising trend until 1990 was also observed in local meteorological data (Shiu et al., 2009). However, the planktonic δ ¹⁸O was consistent overall.

Unlike the relatively continuous planktonic record, the benthic δ ¹⁸O for core 789-L1 was discrete due to insufficient foraminiferal shells at specific layers of the sediment core (Figure 4B). Despite the limited number of measured shells, δ ¹⁸O values of Uvigerina sp. from core 789-L1 were generally lighter than that of core 791-K38. Mulitza et al. (2003) has described that the increase in δ ¹⁸O is induced by the temperature drop, regardless of salinity effect. Therefore, the offset in benthic δ ¹⁸O records (2.2‰ for 789-L1 vs. 2.9‰ for 791-K38 in average between 1910–1940) could be attributed to the different water temperatures between two coring sites (5.51°C at 911 m for 789-L1 vs. 3.45°C at 1,260 m for 791-K38 according to the hydrographic data). However, the variability of the regional upwelling should be another potential factor to change benthic δ ¹⁸O (Wang et al., 2008).

Generally speaking, the planktonic δ ¹⁸O corresponded well with the change in the local surface temperature, and benthic δ ¹⁸O values were controlled by the temperature gradient at the sampled water depth or the regional ocean circulation. There is no further evidence indicating that human activities influenced δ ¹⁸O values in marine sediments in the nearshore realm.

The carbon isotope composition of G. sacculifer of two selected cores are shown in Figure 4C. Unlike the coherent pattern in planktonic δ ¹⁸O records (Figure 4A), the time-series of planktonic δ ¹³C showed a 1‰–1.5‰ decline trend for the last century. Particularly, the decline trend since 1960 is a pervasive feature showing in all collected sediment cores (Figure 3B), which also followed the depletion trend of δ ¹³C in the atmospheric CO₂ (Figure 4H). Since the planktonic δ ¹⁸O did not present a corresponding decline (or incline) trend during the same period (Figures 3A and 4A), the depletion of the planktonic δ ¹³C was not attributed to the temperature change (Goericke and Fry, 1994; Dixit et al., 2015). The results highlight the creditability of foraminiferal records for identifying the ¹³C Suess effect in the nearshore off southwestern Taiwan.

The decline trend of δ ¹³C in the surface ocean was caused by the input of radiocarbon-dead or anthropogenic-produced ¹³C-depleted carbons from the atmosphere through the air–sea exchange process (Suess, 1955; Keeling, 1979; Broecker and Maier-Reimer, 1992; Quay et al., 1992). The ¹³C-depleted carbon has been widely reported in coral skeleton and sclerosponge (Druffel and Benavides, 1986; Swart et al., 1996a; Swart et al., 1996b; Swart et al., 2010; Al-Rousan and Felis, 2013; Hou et al., 2019; Liu et al., 2021) and planktonic foraminifera (Al-Rousan et al., 2004; Black et al., 2011; Xu et al., 2014; Mellon et al., 2019; Simon et al., 2020), while the downcore variability of the benthic δ ¹³C was rather flat and restricted within 0.51‰–1‰ in our study area (Figure 4D). This implies that the penetration of the anthropogenic carbon only occurred in the upper water column at our study site. Although other processes could influence the δ ¹³C of benthic foraminifera (Mccorkle et al., 1990; Schmittner et al., 2017), our results were consistent as the previous finding by Sabine et al. (2004).

The similar magnitude of planktonic δ ¹³C decline from the 1900s were around 0.6‰ and 0.9‰ for cores 789-L1 and 791-K38, respectively (the decline rate for each core was shown as the slope in Table 2). The similar range in the decline of δ ¹³C was first noted by corals and sclerosponges with 0.5‰ between 1850 and 1975 from Bermuda (Nozaki et al., 1978) and Jamaica (Druffel and Benavides, 1986). The global average rate of change in the coral skeletal δ ¹³C was estimated as −0.01‰ year⁻¹ from 1900 to 1990 based on a compilation of coral records throughout the oceans (Swart et al., 2010). However, the decline rates of δ ¹³C in the Indian, Pacific, and Atlantic Oceans were different during 1960–1990. The rate in the Atlantic Ocean was −0.019‰ year⁻¹, but in the Pacific and the Indian Ocean, it was around −0.007‰ year⁻¹ (Swart et al., 2010). The differences could be caused by physiological activities of corals or bathymetric conditions, which change the δ ¹³C_DIC [dissolved inorganic carbon (DIC)] in the ambient seawater (Swart et al., 2010; Watanabe et al., 2017; Fujii et al., 2020; Simon et al., 2020).

To compare with previous coral records (Swart et al., 2010), the δ ¹³C used in this study was analyzed by the linear regression at two corresponding periods (1900 to the present and 1960 to the present, Tables 2 and 3). More than half of the sediment cores used in this study show the statistical significance in the depletion of δ ¹³C (p < 0.05) toward the present (Table 2). From 1900 to the present, the average decline rate of the δ ¹³C in planktonic foraminiferal records was about 0.008‰ year⁻¹, which was similar to the findings of the global coral. From 1960 to the present, the average decline rate of the δ ¹³C significantly increased to 0.016‰ year⁻¹, which was higher than published coral records in the Pacific Ocean but was close to the value of coral measurements in the Atlantic Ocean (Swart et al., 2010). Although the decline rate was different from the previous record in the Pacific Ocean, our findings suggest that the δ ¹³C Suess effect at the surface water of the Pacific Ocean might be as strong as that of the Atlantic Ocean, which is consistent with the global estimation by Eide et al. (2017).

Although there were isotopic offsets between precipitated calcium carbonate shells and the ambient seawater, Mellon et al. (2019) have indicated that decline trends of the δ ¹³C recorded by different species of foraminiferal were still consistent. To compare δ ¹³C among different planktonic foraminiferal records, data published in previous studies were digitalized and plotted with the long-term variability of measured planktonic δ ¹³C in this study (Figure 4E). Our measurement was depicted by LOESS fit (Mellon et al., 2019) and showed a similar trend to planktonic foraminiferal records in the lower latitudes of the Atlantic, both in terms of the overall decline magnitude in, and the enhanced decline rate after, the 1960s. However, there were differences in the δ ¹³C trend found in the higher latitudes of the Atlantic. Our planktonic records showed a larger decreasing range of δ ¹³C than that found in mid to high latitude of the Atlantic, though the rapid decline trend also occurred after the 1960s. The differences have been described by Eide et al. (2017) indicating that the ¹³C Suess effect in the surface seawater is relatively uniform in the North Pacific and North Atlantic but slightly lower in the mid-high latitudes. In the Atlantic, the decline range of δ ¹³C in foraminiferal records was less than 0.3‰ (−0.007‰ year⁻¹) in high latitude, about 0.4‰ (−0.010‰ year⁻¹) in the middle latitude, and close to 0.7‰ (−0.018‰ year⁻¹) in the low latitude (Black et al., 2011; Mellon et al., 2019; Simon et al., 2020). The lower ¹³C Suess effect occurring in the higher latitudes of the Atlantic was attributed to local ocean circulations and primary productivities (Mellon et al., 2019; Simon et al., 2020).

The collected foraminiferal and coral records in sediment cores were further compared with the coral records obtained at Liuqiu (high human impact and close to the coring sites of foraminiferal records) and Dongsha (less human impact). The δ ¹³C trends in foraminiferal and coral records were consistent, and its decline rates were similar (Figures 4C, F–H). The slopes of the Liuqiu and Dongsha coral records after the 1960s were −0.022 and −0.031‰ year⁻¹, respectively (Table 3), which were also higher than that in the published records of Pacific corals. Obviously, the similar δ ¹³C trend found in foraminiferal and coral records was not influenced by the local effect since the distance between Liuqiu and Dongsha is 412 km away (Figure 1), so global ¹³C Suess effect is considered to explain the extensive effect on the records of these two regions. This indicates that planktonic records of sediment cores collected off southwestern Taiwan are suitable to represent the anthropogenic signal of δ ¹³C in the last century.