This study was conducted in the coastal waters of LPMR off the southwestern coast of Puerto Rico ( Figure 1 ), an area designated as a Nature Reserve in September 1979. The reserve consists of a series of reef cays with a dispersed distribution along the interior insular shelf hosting ecosystems, including coral reefs, seagrass meadows, and mangroves (Valdés-Pizzini and Schärer-Umpierre, 2014). Due to the prevalence of south-southeasterly winds, the area is particularly susceptible to SIEs (Hernández et al., 2022). Meteorologically, the LPMR basin is characterized by a wet season extending from August to November, and semiarid conditions prevail during the rest of the year (García-Troche et al., 2021; Ayala-Torres and Otero, 2023). Mangrove litterfall and seagrasses are the major organic carbon sources in LPMR. Net carbon production by planktonic autotrophs was not included as a source of POC in LPMR basin, as the only published information available (Odum et al., 1959) reports net autotrophy presumably supported by dissolved organic carbon (DOC) exported by mangrove forests. Moreover, Meléndez et al. (2022), using data from a decade of observations collected by La Parguera MapCO₂ buoy, located off a mid-shelf reef key in the LPMR basin, reported net heterotrophic conditions during the year as slightly autotrophic conditions only prevailing during winter months.

Six (6) Sargassum traps ( Supplementary Figure A.1 in Supplementary Materials ), constructed using PVC pipe and plastic mesh and measuring 0.63 x 0.5 x 0.63 m (depth x width x height), were deployed facing the prevailing southeasterly winds on the seaward edge of four reef islands on the outer southern boundary of LPMR basin. An additional trap was deployed in Monsio Jose Key Bay (MJKB) (17.9688°N, 67.076871°W), a nearshore mangrove-lined embayment ( Figure 1 ). Traps were placed in 10 cm deep water to ensure water inflow even at low tide. Weekly sampling facilitated trap maintenance, allowing for continuous assessment of their condition. Traps were replaced as needed to ensure optimal functionality and uninterrupted sampling.

Quantification of the weekly Sargassum biomass influx (kg·m^-1·Wk^-1) into the LPMR basin and MJKB was estimated by collecting the Sargassum accumulated in the traps, transferring it to a mesh bag, and weighing it on-site using an electronic fish scale. The Sargassum wet weight was converted to POC using an averaged carbon-to-wet weight ratio of 0.05 ± 0.01 published by Laffoley et al. (2014); Wang et al. (2018, 2019); and Gouvêa et al. (2020) ( Supplementary Table A.1 in Supplementary Materials ). The total weekly Sargassum carbon influx to the LPMR basin was estimated using the weekly mean capture of all traps located in the outer reefs, normalized by the trap width (meters) and multiplied by the width of the basin’s windward boundary (10.4 km). For MJKB, weekly mean values were multiplied by the width of the channel (69 m) facing the prevailing wind. Sargassum carbon influx rates to MJKB nearshore station were contrasted with estimates of carbon influx from mangrove litterfall. To achieve a more comprehensive assessment of the carbon contribution, we also estimated DOC from Sargassum using values reported by Powers et al. (2019).

The mangrove litterfall rate for the LPMR basin and MJKB were estimated using the mean of reported litterfall observations in LPMR, 476 dry weights·m^-2·yr^-1 (Vega-Rodríguez et al., 2008; Pérez-Pérez et al., 2022) and area estimates were derived from satellite imagery. Mangrove litter mass was converted to carbon using the 0.5 carbon/litter weight ratio reported by Golley et al. (1962) ( Supplementary Table A.2 in Supplementary Materials ). The estimated POC was converted to DOC using the 0.13 reported by Adame and Lovelock (2011).

Estimates of net carbon input from seagrasses are based on seagrass growth studies carried out in LPMR basin by Liboy (1976) and Hertler (2002). The mean seagrass productivity rate (4.56 ± 2.01 g·m^-2·day^-1) calculated from data from both studies was used to obtain the seagrass POC production rate for the basin. Said rate is consistent with reports from other areas in the Caribbean (Linton and Fisher, 2004; Juman, 2005). Seagrass biomass was converted to carbon using the carbon-to-biomass ratio (0.32) reported by Bay et al. (1996) ( Supplementary Table A.3 in Supplementary Materials ). We estimated the exudation carbon by using the POC-to-DOC ratio (0.126) reported by Robertson et al. (1982).

Although Sargassum traps were deployed throughout LPMR, the analysis of weekly seawater samples for assessing the biogeochemical impact of SIEs was exclusively conducted for MJKB in this study. Samples were collected within 3.5 meters of the MJKB Sargassum trap (17.968766°N, 67.076871°W) from January 2022 to December 2022 between 7:00 and 10:00 a.m. (local time) at 1-meter depth using a Van Dorn 3.5 L sampler, following the best practices guidelines (Dickson et al., 2007). One seawater sample was collected for each parameter, which allowed for duplicate analyses in the lab. Conductivity and temperature data were collected with an SBE25 CTD. Seawater samples for pH and TA were fixed immediately with a saturated solution of mercury chloride (HgCl₂) to prevent biological alteration. Analysis for pH on the Total Scale was performed using a spectrophotometer with m-cresol purple indicator dye (pH_T ± 0.003) (Dickson and Goyet, 1996; Grasshoff et al., 2007). Total alkalinity determinations (TA ± 2 μmol·kg^-1) (Dickson et al., 2007) were carried out following the protocol described by García-Troche et al. (2021). DO sample analysis was performed following the Winkler method (DO ± 0.50 mg·L^-1) (Grasshoff et al., 2007; Astor et al., 2013). Aragonite saturation state (Ω) values were estimated from pH and TA measurements using the CO2SYS program (Lewis and Wallace, 1998).

Pearson’s correlation analysis was used to identify significant time-lagged correlations between the explanatory variable (i.e., weekly Sargassum carbon influx) and the dependent variables (i.e., DO). A MATLAB function was created to identify different weekly lags between the variables and show the significant Pearson’s correlation coefficient. The lagged Sargassum carbon influx (kgC·m^-1) and physical parameters, such as wind speed (m·s^-1) and changes in tide height, calculated as Δh = (tide height at sampling time)/(mean low tide), were included in data analyses to determine their significance in modulating the measured and calculated biogeochemical parameters (i.e., DO, pH, Ω). Wind speed data were sourced from the National Buoy Center, and tidal data were obtained from NOAA Tides & Currents for Station 9759110, Magueyes Islands, PR.

The MATLAB Fathom toolbox (Jones, 2017) was used to perform a stepwise forward selection of explanatory variables in Redundancy Analysis (RDA) using Akaike Information Criteria (AIC). This analysis identified optimal variables that substantially explained the variation of biogeochemical parameters (i.e., response variables; DO, pH, Ω). Explanatory variables included in the RDA-AIC analysis were Sargassum carbon influx, wind speed, and Δh. Subsequently, a permutation-based RDA with 1000 iterations was conducted using the optimal explanatory variables identified through the RDA-AIC analyses to derive the model statistics. Lastly, to gain a clearer understanding of the individual effects of the optimal explanatory variables on the response variables, a permutation-based Multiple Linear Regression via Least Squares Estimation with 1000 iterations were performed for each response variable independently. This enabled a more precise interpretation of the impact of each explanatory variable on the response variables, offering insights into their distinct roles within the broader model.

During 2022, SIEs at LPMR basin started in April and extended until November. The mean weekly Sargassum biomass influx for the six (6) traps, located in the outer keys of the LPMR basin ( Figure 2 ), ranged from non-detectable to a maximum of 24.80 kg·m^-1·Wk^-1 occurring on the second week of May. An estimate of the mean weekly Sargassum biomass influx into the basin yields 7.85 ± 6.60 kg·m^-1·Wk^-1 or 81,725 kg·Wk^-1 for the whole basin. For the same period, the weekly Sargassum biomass influx rate into the MJKB averaged 2.13 ± 5.08 kg·m^-1·Wk^-1 with a maximum Sargassum input rate of 23.22 kg·m^-1·Wk^-1 ( Figure 3 ). Extrapolation using the width of MJKB channel (69 m) aligned with the prevailing wind direction, yields a weekly mean Sargassum biomass loading rate for the embayment of 155 ± 327 kg·Wk^-1.

Figure 4 presents the estimates of POC production by mangroves (as litterfall) and seagrasses (leaf growth) as well as estimates of Sargassum POC influx into the LPMR basin. The weekly POC production of seagrasses and mangroves totaled 18,209 kgC·Wk^-1, while the POC loading arising from the Sargassum influx during high season in 2022 averaged 3,617 ± 3,452 kgC·Wk^-1 with a standard error of 241 kgC·Wk^-1, thus representing about 20 ± 19% increase in POC input to LPMR basin ( Figure 4 ).

The weekly estimated exudates of DOC from the production of seagrasses and mangroves in the LPMR basin are 3,041 kgC·Wk^-1 ( Supplementary Table A.4 in Supplementary Materials ). At the same time, the DOC loading arising from the Sargassum influx during the high season in 2022 averaged 6 kgC·Wk^-1, thus representing a 0.2% minor fraction increase in DOC input to the LPMR basin. The comparison of the calculated POC and DOC fractions from Sargassum suggests that POC is the predominant contributor to the organic carbon pool in the LPMR basin.

Weekly estimates of carbon loading from mangroves (as litterfall) and Sargassum to MJKB are presented in Figure 5 . Nonetheless, for MJKB, the seagrass carbon was not considered because the study site does not harbor seagrasses. While carbon production by mangroves for the area in the MJKB, is estimated at 7.5 ± 1.4 kgC·Wk^-1, Sargassum, POC influx averaged 7.7 ± 16.3 kgC·Wk^-1, thus representing a 103% net increase in POC.

The carbon exudation estimates by mangroves, primarily through litterfall, for MJKB are approximately 1.0 kgC·Wk^-1 ( Supplementary Table A.5 in Supplementary Materials ). In contrast, the DOC influx from Sargassum into the MJKB averaged 0.013 kgC, contributing to a minor increase of 1.3% in DOC.

Observations of biogeochemical data indicate that during the months before the arrival of Sargassum (winter season), pH ranged from 7.7 to 7.9, while Ω _aragonite ranged from 2.2 to 3.0. DO values ranged from 3.28 to 5.24 mg·L^-1, while temperature ranged from 26.11 to 28.21°C ( Supplementary Table A.6 in Supplementary Materials ). After the onset of the Sargassum season in early May, we observed a sharp decrease in pH, Ω_aragonite, and DO ( Figure 6 ). Simultaneously, we observed a warmer seawater temperature. For this period, pH values ranged between 7.0 – 7.8, with increased seawater acidity observed during the summer months when Ω _aragonite ranged between 0.5 – 2.8, values under critical levels are Ω < 2.0. During the same period, DO values ranged from non-detectable to 4.67 mg·L^-1, frequently reaching hypoxic or anoxic conditions. Temperature values ranged from 27.90 to 30.77°C, with higher temperature levels occurring between late summer and fall. The ecosystem’s gradual and modest recovery is evident towards the end of the season, albeit with sustained low DO and pH levels. These conditions persisted from mid-June to September.

Data analyses were performed using a one-week lag on the Sargassum carbon influx based on the significant time lagged correlations identified by the Pearson’s correlation analysis using DO, pH and Ω data ( Supplementary Table A.7 in Supplementary Materials ). These results suggest that biogeochemical parameters exhibit measurable changes one week after a SIE, indicating Sargassum impact on the system’s chemical dynamics. The stepwise RDA-AIC analyses showed that Δh and Sargassum carbon influx were the optimal explanatory variables for the variance of DO, pH and Ω ( Table 1 ). A significant portion of the response variables is explained by Δh independently, but adding Sargassum carbon influx further improves the model. Wind speed was not identified as an optimal explanatory variable by the AIC analyses. The RDA permutation test demonstrated that the model incorporating the optimal explanatory variables identified through RDA-AIC accounted for a significant proportion of the variance in the response variables (p < 0.05, r² = 0.34; Supplementary Table A.8 in Supplementary Materials ).

Multiple linear regressions to assess the individual effects of Δh and Sargassum carbon influx on response variables (i.e., DO, pH, Ω) showed that DO is significantly influenced by Δh and Sargassum carbon influx (p < 0.05, r² = 0.35; Table 2 ), with Sargassum carbon influx having a negative relationship and Δh having a positive relationship with DO. However, the response of pH and Ω is less well explained by Δh and Sargassum carbon influx (r² = 0.14). The variation in Δh has a marginal influence on pH, whereas the influx of carbon from Sargassum has a minimal impact on Ω ( Table 2 ).