Influence of local industrial changes on reef coral calcification.

Coral reefs are currently facing multiple disturbances caused by natural/anthropogenic factors. Recent industrial development might influence reef environments and ecosystems; however, few direct comparisons of coral calcification with the histories of local industries exist. We show the coral Ba/Ca record and growth histories for 46 years collected from Sumiyo Bay, Amami-Oshima Island, Japan. Coral Ba/Ca was mainly controlled by the sediment loads in seawater, which are introduced through the two local rivers. Coral Ba/Ca records have been characterized by two distinct historical periods: the decadal fluctuation corresponding to the traditional silk fabric industry (1960s ~ 1995) and the increasing trend corresponding to the development of quarries and the construction industry (1996 ~). Coral Ba/Ca records and local industrial histories were also linked to coral calcification. A long-term quantitative assessment of reef environments and local industrial changes could provide an evaluation of the survival strategies of reef-building corals in the future.

. Regional map around the sample site (Sumiyo Bay, Amami-Oshima Island). Sumiyo River and Yakugachi River flow into Sumiyo Bay. The solid line area indicates a mangrove forest located at the rivers' mouth, and the broken line area shows quarries in the southern coastal area of Sumiyo Bay. The white doublecircle shows the site where Porites coral was drilled in this study. The distance between the rivers' mouth and the sampling site was 4.5 km. The white single circles with a number show the site where the water samples were collected (No. 1, 2: Sumiyo River; No. 3: mangrove forest; No. 4, 5: Yakugachi River; No. 6, 7: Sumiyo Bay (brackish water); and No. 8,9,10: Sumiyo Bay (seawater)). The map figures (Japan, Amami-Oshima Island, and Kikai Island) were generated using Generic Mapping Tools (GMT v. 5.3.3) 59 . Satellite image (the Sumiyo area) was obtained by Google Maps (https://www.google.co.jp/maps/); Imagery ©2020 Maxar Technologies, Maxar Technologies, Data SIO, NOAA, U.S. Navy, NGA, GEBCO, TerraMetrics, Imagery ©2020 Landsat/Copernicus, Maxar Technologies, Map data ©2020. These figures were modified using Microsoft ® PowerPoint for Mac (v. 16.35).

Results and Discussions
Ba/Ca profiles in the porites coral and Ba concentration in river-and sea-waters. The Porites coral specimen showed 46 clear annual density bands ( Supplementary Fig. 1). We analyzed skeletal Sr/Ca and skeletal Ba/Ca records for 46 years from 1968 to 2014 ( Supplementary Fig. 2a,f). The 46-year skeletal Ba/Ca record derived from the Sumiyo coral showed multidecadal fluctuations, which seemed to be larger than seasonal variability. Multidecadal fluctuation in the skeletal Ba/Ca record was isolated by Gaussian bandpass filter (i.e., low-pass filtered skeletal Ba/Ca; Fig. 2b, see also Materials and Methods). The detrended skeletal Ba/Ca showed spike-like patterns (Fig. 2c).
The coral skeletal Ba/Ca reflects seawater Ba concentration (Ba/Ca coral /Ba/Ca seawater ≒ 1) 18 . Our data on the Ba concentration in seawater ranged from 32.1 to 45.1 nmol/kg ( Supplementary Fig. 3). The mean value of measured Ba concentration in seawater (39.0 nmol/kg) mostly agreed with the variable range of the coral-derived Ba concentration in seawater in Sumiyo Bay ( Supplementary Fig. 4), and it was slightly higher than the reported values in previous studies (Supplementary Table 1). There was a strong negative correlation between the Ba concentration in the water samples and the in situ salinity in the Sumiyo region ( Supplementary Fig. 3, r = −0.89; p < 0.0001; n = 14). Ba is released/desorbed from particulate matter by ion exchange with seawater Mg 2+ and Ca 2+ ions in a low-salinity estuary 19,20 . The Ba concentration in seawater in Sumiyo Bay is controlled by the amount of the desorbed Ba 2+ (in the estuary) from the Ba-containing sediment, which is transported from the two rivers through the mangrove forests with tides and precipitation effects. The primary source of Ba in the estuary is suspended fluvial sediment particles, which are transported by rivers towards the ocean with freshwater/ flood plumes [19][20][21] . The river flood event would lead to an increase in the river end-member Ba concentration 21 . The desorbed Ba 2+ from the particulate matter is readily substituted for Ca 2+ in the coral aragonite lattice in proportion to the aqueous Ba/Ca ratio 18,22,23 . Multidecadal fluctuation in skeletal Ba/Ca. The trend of the low-pass filtered skeletal Ba/Ca was analyzed by SiZer trend analysis 24 (see Materials and Methods). We found a trend shift point in 1995 ~ 1996, which is the significant increase trend changed into the significant decrease trend in <25 years of filter width ( Supplementary Fig. 5). Thus, we divided the time series into two periods in this study: 1986 ~ 1995 (A-period) and 1996 ~ 2014 (B-period). Ba/Ca in corals might be controlled by following scenarios: (1) Sea surface temperature (SST) components of Ba/Ca 22,25 , (2) Ba uptake in sea surface due to upwelling 22,26,27 , and (3) Ba-containing sediments transported from the river (freshwater/flood plumes) [11][12][13][14][15][16] . For our skeletal geochemical record, the annual mean of low-pass filtered skeletal Ba/Ca was not correlated with low-pass filtered skeletal Sr/Ca ( Supplementary Fig. 6a, r = −0.17; p = 0.26; n = 47), and it suggested that the skeletal Ba/Ca did not show similar variations with SST. Therefore, the skeletal Ba/Ca has not been affected by SST on the multidecadal scale. Moreover, the upwelling effect in this study was very small because the influence of the Ryukyu Current System was not remarkable and the correlation coefficient value (r) between the low-pass filtered skeletal Ba/Ca and the low-pass filtered coral derived δ 18 O seawater was low ( Supplementary Fig. 6b, r = 0.29; p < 0.05; n = 47, see also Supplementary text). Thus, there was the other control factor of the annual mean of low-pass filtered skeletal Ba/ Ca in Sumiyo Bay, for example, natural/anthropogenic sediment load.
Ba/Ca in corals reflects anthropogenic sediment load in the coastal area (e.g., land-use, coastal development) 11,13,15 . The industrial development around Sumiyo Bay in the past 46 years 17,28,29 can also be divided into two periods (A-period and B-period, Fig. 2a), and it was agreed with the significant trends shift in the low-pass filtered skeletal Ba/Ca. In the A-period, from the 1960s to 1995, the production of "Oshima tsumugi" and forestry prospered 17,28 . Oshima tsumugi is a traditional silk fabric on Amami-Oshima Island. It is made by "mud dyeing" with plant-derived tannic acid pigments and iron oxide contained in mud fields. In the A-period, there were many mud fields for dyeing along rivers, and the process of mud washing-off was carried out in the rivers 17 . Forestry in the A-period produced a substantial amount of pulp and chips (a timber factory in Sumiyo village closed in 1995) 17,28,29 . In the B-period, from 1996 to the present, the quarry and the construction industries (e.g., public works such as construction and maintenance work on roads, rivers, and the bay) prospered 17,28,29 . In addition, the farming (e.g., citrus cultivation) also prospered during both periods 17,28,29 . These local industries might have influenced the reef environment in Sumiyo Bay. It is necessary to compare the local archives on the change in the thriving industries with coral calcification responses to understand the anthropogenic reef disturbance.
The low-pass filtered skeletal Ba/Ca record and the annual production of Oshima tsumugi showed similar patterns (Fig. 2b, see also Materials and Methods). There was a significant positive relationship between the low-pass filtered skeletal Ba/Ca and the production volume of Oshima tsumugi (Fig. 3, r = 0.86; p < 0.0001; n = 26; from 1968 to 1993). This result suggests that the multidecadal fluctuation in the skeletal Ba/Ca in the A-period reflected the mud or sediment load in Sumiyo Bay due to "mud dyeing" (including the silk thread dyeing at the muddy fields, the mud washing-off process carried out in the rivers, and then the mud was transported to the ocean by the rivers). In addition, deforestation in the A-period also could have affected the low-pass filtered skeletal Ba/ Ca. In the A-period, the sediment or mud may have constantly been flowing out into Sumiyo Bay due to forestry and Oshima tsumugi production. On the other hand, in the B-period, the low-pass filtered skeletal Ba/Ca had an upward pattern (Fig. 2b). The multidecadal fluctuation in the skeletal Ba/Ca in the B-period might have reflected an increase in sediment load due to rapid industrial development. In the B-period, a large amount of sediment flowed into Sumiyo Bay, which would be caused by the construction industry and quarries. In addition, the quarry collapse events (October 2011; November 2011; March 2014) 30 would cause sediment input sporadically into Sumiyo Bay (Fig. 2e).
Detrended skeletal Ba/Ca. The detrended skeletal Ba/Ca record was characterized by spike-like patterns (i.e., high-peaks). Skeletal Ba/Ca high-peaks were recorded 43 times in the A-period and 33 times in the B-period (Fig. 2c). For the river flooding disasters associated with typhoons or heavy rainfall events 17,30,31 , the skeletal Ba/  17,28,29 . In the B-period (1996 ~ 2014), the construction industry and quarries prospered 17,28,29 . Farming also prospered during both periods 17,28,29 . (b) Annual mean of low-pass filtered skeletal Ba/Ca (in black) compared with the production of Oshima tsumugi (in blue). The low-pass filtered skeletal Ba/Ca was isolated by Gaussian bandpass filter (frequency = 0; bandwidth = 0.00137) 53 from the raw skeletal Ba/Ca data, which is shown in Supplementary Fig. 2f (see also Materials and Methods). The standard unit of the production of Oshima tsumugi is "Tan" (in Japanese), which is defined as approximately twelve meters long and thirty-seven centimeters wide. (c-e) Detrended skeletal Ba/Ca compared with monthly precipitation. The detrended skeletal Ba/Ca was obtained by Gaussian bandpass filter (frequency = 0; bandwidth = 0.00137), as notch filtered data 53 . The gray line on the detrended skeletal Ba/Ca record indicates the mean seasonal cycles (MSC). The monthly color circles show the skeletal Ba/Ca high-peaks, which determined the difference between the mean value of MSC and the detrended data (see Methods and Materials). The blue and red precipitation areas show typhoons or heavy rainfall events (blue) and river flood disasters corresponding to typhoons or heavy rainfall (red) 17 www.nature.com/scientificreports www.nature.com/scientificreports/ Ca high-peaks were recorded in September 1977 (with a 3-month time lag from the event), September 1990 (without a time lag), October 1991 (with a 2-month time lag), and February 2011 (with a 4-month time lag). The high-peak values in response to these events in the B-period were larger than those in the A-period (Fig. 2d,e). It is possible that the response of the coral skeleton to these events has changed due to local industrial development, throughout the two periods.
There are two possible reasons for the time lag of several months that occurred between the river flooding events associated with heavy rainfall events and the peaks of the skeletal Ba/Ca: (1) a delay in the desorption of Ba 2+ in the mangroves and estuary and (2) a decrease in or cessation of skeletal growth (for the 2010 event). (1) The estuarine Ba 2+ desorption processes that affect the Ba concentration in seawater are known to be complex. Moyer et al. 32 noted a time lag of several months between the peak of skeletal Ba/Ca and the peak of the river discharge, which was explained by the sporadic release of Ba from estuarine or floodplain-stored sediments. However, a large amount of sediment and freshwater will flow into the bay rapidly and suddenly when the limit of the mangrove trapping mechanism is exceeded due to heavy rainfall, river floods, and mangrove collapses. A large amount of sediment that flows out without being trapped by mangroves will be stored in the floodplains and estuaries, and then, the desorption of Ba 2+ becomes sporadic and slow due to tidal mixing. As a result, the increase in the Ba concentration in Sumiyo Bay will be delayed. (2) For the heavy rainfall disaster that occurred in October 2010, the daily precipitation of that day (622.0 mm) was highest in the 46 years past. The 2010 heavy rainfall specifically triggered a river flood and the collapse of mangrove forests, which caused a large amount of sediment and freshwater input into Sumiyo Bay 30,31 . The skeletal Ba/Ca record tracked the changes in sediment load due to the 2010 disaster showing an increasing trend for 4 months, and a maximum value of skeletal Ba/Ca was 5.32 μmol/ mol (1.4 times the mean value over all periods). Similarly, the skeletal Ba/Ca record showed a time-lag signal that rose over several months during the heavy rainfall event that occurred in September and November 2011. Clear low-density bands were observed in the X-radiographs from 2010 to 2012 when the coral experienced heavy rainfall, river floods, and landslide events (Supplementary Fig. 1). Anomalous high-/low-density bands are skeletal growth responses to environmental stressors, for example, thermal stress, sedimentation, or substantial precipitation 33 . In addition, the skeletal growth disturbance proxy, ΔSr/Ca in the Sumiyo coral, showed high-peaks after the 2010 event ( Supplementary Fig. 2g). The skeletal Sr/Ca and Mg/Ca in the Sumiyo coral recorded a low sea surface temperature (SST, 17.6~18.6 °C derived by SST-Sr/Ca and SST-Mg/Ca regression; Supplementary  Figs. 2a,b and 7, see also Supplementary text) after the 2010 river flood event. This scenario suggests that cold freshwater flowed into Sumiyo Bay from the Sumiyo and Yakugachi Rivers in the aftermath of the heavy rainfall event. The temporal heavy rainfall also probably influenced on corals. Fallon et al. 34 explained that the time lag of the skeletal trace element record was caused by the temporal decrease in or cessation of skeletal growth during winters with lower SST than 18 °C. Thus, skeletal growth was temporarily inhibited by the 2010 river flood event, which might have created a time lag. . There was no significant difference between the two periods for monthly precipitation and SST.

Seasonal pattern in skeletal Ba/Ca.
Only during the typhoon and heavy rainfall season (August to October) was the monthly mean of skeletal Ba/Ca showed a similar pattern with precipitation (A-period: r = 0.99; p < 0.05; n = 3, B-period: r = 0.96; p = 0.19; n = 3, Fig. 4). www.nature.com/scientificreports www.nature.com/scientificreports/ This result conflicts with the simple hypothesis that Ba-rich sediment particles are transported towards Sumiyo Bay by rivers with freshwater plumes associated with precipitation throughout the year. The estuaries serve as storage reservoirs for Ba-rich sediments [35][36][37] . The mangrove forest is one of the storage reservoirs. The mangrove can trap suspended sediments that are caused by river runoff or resuspension of bottom sediments [38][39][40] . The trapping mechanism has a limitation in which too much sedimentation can lead to mangrove mortality as the sediments asphyxiate the respiratory structures 41 . In addition, the mangroves are affected by both river (river runoff related to precipitation) and ocean effects (seawater inflow and outflow due to tide levels). Increased river flow due to precipitation makes sediment transport more active and leads to sediment inflow into mangroves 42 . Heavy rainfall and strong winds associated with typhoons or storms sometimes trigger river floods and landslides, which can cause the outflow of large amounts of sediments into the ocean with mangrove collapse 43 . Seawater flows into the mangroves during the springtide and flows out during the ebb tide 40,44,45 . Therefore, sediment transport is most active during springtide. The tide level in our study area was higher in summer and autumn (June to November) than in winter and spring (December to May). From August to October, typhoons and heavy rainfall events and river floods frequently occurred (Fig. 2c). Therefore, Ba-containing sediments might slowly flow into Sumiyo Bay from mangroves during winter to early summer (November to June, when there might be no influence of typhoons or heavy rainfall). On the other hand, the sediments exceeding the limit of the mangrove trapping mechanism might suddenly flow into Sumiyo Bay during summer to autumn (July to October, when typhoons, heavy rainfalls, and river floods occur frequently). From May to June, the sediment input to Sumiyo Bay is controlled by the mangrove trapping mechanism, although there is substantial precipitation. In addition, the low-pass filtered skeletal Ba/Ca (i.e., multidecadal fluctuation) did not show the SST effect; however, our skeletal Ba/Ca may have been influenced by SST on the seasonal scale. This finding is consistent with that of previous studies 22,25 . In particular, during winter to spring (December to May), the skeletal Ba/Ca may reflect the change in SST because the tide level is low and sediment transport from mangrove forests is considered to be less active. Thus, the skeletal Ba/Ca in the Sumiyo coral may reflect the mangrove system (affected by both tide level and precipitation) and SST variation on the seasonal scale. Moreover, the sensitivity (response) of the reef-building coral to sediment load has changed with the change in the rhythm of sediment outflow into Sumiyo Bay throughout the two periods. The results of the difference in the variable range of the seasonal trend between the two periods and the increase in spike-like signals in the B-period can be explained by the histories of the local industries.
Coral calcification response to natural/anthropogenic sediment load. Coral skeletal growth parameters (skeletal extension rate, skeletal density, and skeletal calcification rate) are obtained from annual bands and age models (see Materials and Methods). As shown in Fig. 5, there were the significant relationships between the annual mean of the low-pass filtered skeletal Ba/Ca and the annual extension rate in both periods (A-period: r = −0.53; p < 0.01; n = 27; B-period: r = −0.62; p < 0.01; n = 18). Besides, there were also the significant relationships between the low-pass filtered skeletal Ba/Ca and the annual calcification rate in both periods (A-period: r = −0.55; p < 0.01; n = 27; B-period: r = −0.76; p < 0.001; n = 18). Therefore, natural/anthropogenic sediment load was likely one of the controlling factors of coral calcification in Sumiyo Bay. Skeletal calcification is enhanced by light availability (i.e., inhabitation of light-enhanced calcification 46 ). Numerous previous studies have reported that increasing the sediment load in seawater (i.e., high water turbidity) due to coastal runoff cause would reduce the coral light availability, and it will lead to disturbances in coral calcification [47][48][49] . Besides, the correlation coefficient r-value of skeletal Ba/Ca vs. annual calcification rate in the B-period (r = −0.76; slope: −0.36 ± 0.08) was larger than that in the A-period (r = −0.55; slope: −0.29 ± 0.09). There is a possibility that coral calcification in the B-period could more strongly respond to natural/anthropogenic sediment load than in the A-period.

Conclusion
The skeletal Ba/Ca derived from the Sumiyo Bay coral was mainly controlled by the amount of the desorbed Ba 2+ (in the estuary) from the Ba-containing sediment transported by the two rivers and mangrove forests. The tides and precipitation effects are shown as the seasonal pattern in skeletal Ba/Ca. Our skeletal Ba/Ca records were characterized by two distinct periods: (A) the multidecadal change with smaller seasonal variation during the 1960s to 1995 corresponding to the change in the traditional silk fabric industry and (B) the increase pattern with larger seasonal variability during 1996 to the present corresponding to the development of quarries and the construction industry. Thus, the sensitivity (response) of the reef-building coral to sediment load has changed with the change in the rhythm of sediment outflow into Sumiyo Bay throughout the two periods. Moreover, we concluded that natural (the sedimentation trapping mechanism in the mangroves, precipitation effects, and river flooding effects) and anthropogenic sediment loads (industrial development and land-use effect) were two of the controlling factors of coral calcification. www.nature.com/scientificreports www.nature.com/scientificreports/ The coral growth record showed a response to coastal environmental change in the Sumiyo area. It indicates that Sumiyo corals can adapt to environmental changes (i.e., chronic sediment stress) if it is a gradual change over a long period (>approximately 50 years). The multiple and mixed components of natural and anthropogenic factors (i.e., human activity, climate change, and land-use change) will influence coral reefs 1,2,4,5,9,10 . A long-term quantitative assessment can simultaneously evaluate a reef environment and coral calcification. It can be essential to the discussion of survival strategies for reef-building corals in which precipitation and human activities will increase. If a sudden and larger environmental change than the natural/anthropogenic events reconstructed in this study occurs consecutively, it might have a stronger negative impact on the reefs. It is important to understand how natural/anthropogenic sediment loads lead to a reef disturbance and how corals respond to these loads to create a sustainable human society that can benefit from coral reefs. In addition, keeping the balance between natural/anthropogenic reef disturbance and the coral calcification response may also be needed to maintain reef health and reef biodiversity in the future. There are still many coastal areas where periodic monitoring surveys and in situ environmental data are not available. Massive coral skeletons (e.g., Porites sp.) can reconstruct reef disturbances with a high temporal resolution, and these skeletons will be a useful proxy for skeletal growth and calcification responses even in such coastal areas.

Materials and Methods
Coral core collection and geochemical analysis. We collected living Porites coral cores at a depth of 5.1 m inside Sumiyo Bay (28.23N 129.45E, Fig. 1) in October 2014. The distance between the mouths of the two rivers (Sumiyo River and Yakugachi River) and the coral sampling site was 4.5 km, and the distance from the nearest shore was 80 m. All cores were cut into 5 mm thick slabs. The slabs were rinsed with distilled water in an ultrasonic bath several times and dried at 50 °C for 24 h. X-ray photographs were taken at 50 kV and 1.0 mA for 8 seconds using a digital X-ray scanner (NAOMI, RF Co., Ltd, Japan) to observe annual density bands ( Supplementary Fig. 1). Analytical lines were determined along the major growth axis. The slab edges formed into a ledge shape, 1.5 mm thick and 2.0 mm width along the analytical lines, using a micro drill. Skeletal powder samples for geochemical analysis were obtained by a precise microsampling method using a micro drill (1.8 mm diameter) and a PC controlled XY stage, as the method described in Watanabe et al. 50 . Microsampling intervals were 0.25 mm, and it was equivalent to 0.3 months resolution (the mean value of the annual skeletal extension rate was 9.6 mm/year). The milled area was cleaned by high-pressure air after each microsampling to avoid cross-contamination.
The skeletal Ba/Ca, Sr/Ca, and Mg/Ca analyses were performed by ICP-OES (iCAP 6200, Thermo Scientific, MA, USA) with an autosampler (ASX-260, Teledyne CETAC Technologies, NE, USA) and an ultrasonic nebulizer (U-5000AT, Teledyne CETAC Technologies). Powdered samples (115-128 μg) were dissolved in 0.57 ml of 25% high-purity HNO 3 and then diluted with ultrapure water (Milli-Q), as described by Watanabe et al. 51 . We ultimately prepared sample solutions in which the Ca concentration was 9 ppm. Calibrations of the five gravimetric standard solutions yielded high correlation coefficients (r 2 ) of 0.9999 for Ba, Sr, Mg, and Ca. To correct the instrumental drift, we measured the reference solution JCp-1 52 at 5 sample intervals. The analytical uncertainties (1σ; relative standard deviations) were 2.24% or 0.18 μmol/mol for Ba/Ca, 0.17% or 0.01 mmol/mol for Sr/Ca, and 0.73% or 0.03 mol/mol for Mg/Ca. These profiles are shown in Supplementary Figs. 2a (Sr/Ca), 2b (Mg/Ca), and 2f (Ba/Ca).
The skeletal δ 18 O and δ 13 C analyses were performed using a mass spectrometer (Finnigan MAT 253, Thermo Scientific) coupled to a carbonate reaction device (Kiel II and IV Carbonate Device, Thermo Scientific). Powdered samples (100 ± 10 μg) were reacted with phosphoric acid at 70 °C in a carbonate reaction device. The CO 2 gas produced in the reaction was then introduced into a coupled mass spectrometer. Skeletal δ 18 O and δ 13 C values are expressed in conventional delta notation in ‰ units relative to Vienna Peedee Belemnite (V-PDB) through measurements of the isotopic ratio of CO 2 gas derived from Standards NBS-19 (δ 18 O = −2.20‰; δ 13 C = 1.95‰). The standard deviations (2σ; 95% confidence) were 0.02‰ and 0.03‰ for δ 18 O and δ 13 C, respectively. These profiles are shown in Supplementary Figs. 2c (δ 18 O) and 2d (δ 13 C).
Statistical analysis. The skeletal Ba/Ca record was filtered with a Gaussian bandpass filter (frequency = 0; bandwidth = 0.00137, i.e., isolating >2 years or <2 years trend) using Analyseries 2.0.8 53 . The profiles of raw data and low-pass filtered data (isolating >2 years trend) are shown in Supplementary Fig. 2f. The detrended data (i.e., notch filtered data, isolating <2 years trend) are shown in Fig. 2c. The core-top data that showed anomalously high values (i.e., tissue layer, 4.1 mm = 5.1 months) were excluded from the filtering analysis. The high-peaks in the low-pass filtered skeletal Ba/Ca were determined the difference between the mean value of mean seasonal cycles (MSC) and the detrended data. When the peak points succeeded one another, the highest point was determined as a representative point (i.e., high-peak) in Fig. 2c-e. Moreover, the time series of the low-pass filtered skeletal Ba/Ca was divided into two periods using SiZer (SIgnificant ZERo crossings of derivatives) method 24 . The analysis was performed using the statistical software R (v. 3.3.3) 54 with Significant Zero Crossings for R package (v. 0.1-4) 55 . The package was obtained from https://CRAN.R-project.org/package=SiZer. As a result, there is a significant trend shift point in 1995~1996 (Supplementary Fig. 5). We divided the time series of the low-pass filtered skeletal Ba/Ca using the shift point (A-period: 1968-1995, B-period: 1996-2014). Because the increase and decrease trends were overlapped from 1975 to 1978 in a broad range of the filter width, we did not divide the time series of the low-pass filtered skeletal Ba/Ca. River-and sea-water sample collection and analysis. We collected 14 water samples in Sumiyo town (Fig. 1). Two of them were collected from the coral core drilling site (inside Sumiyo Bay) in October 2014. The other samples were collected from Sumiyo River, Yakugachi River, the mangrove forest, and Sumiyo Bay in July