Extreme rainstorms drive exceptional organic carbon export from forested humid-tropical rivers in Puerto Rico

Rivers play multiple roles in the geological carbon cycle, acting as net atmospheric CO2 sinks in some catchments (carbon- and sulfide-poor bedrock) and sources in others (abundant petrogenic organic carbon or sulfides). Biogenic organic carbon (OC) erosion and burial, and dissolved inorganic carbon (DIC) production during silicate weathering, are central to riverine contributions to long-term carbon budgets. Globally, river biogenic particulate organic carbon (POC) export is commonly inferred from suspended sediment (SS) yields via empirical regressions; however, most compiled rivers drain sedimentary basins rich in petrogenic carbon. In humid-tropical igneous montane islands and island-arc terrains, petrogenic carbon and sulfides are scarce, potentially making these systems strong carbon sinks, but they have been under-assessed. POC export is highly event-driven and rarely sampled during extreme flows, despite projections that extreme tropical rainfall will intensify. This study quantifies SS and carbon (biogenic POC, DOC, DIC) yields over 25 years, emphasizing hurricanes and major storms, in two humid-tropical, forested, petrogenic-carbon-poor catchments (Icacos, granodiorite; Mameyes, volcaniclastic) in Puerto Rico. We estimate yields, coastal burial, contributions to the geological carbon cycle, NPP export as POC, event-scale flux partitioning, and compare SS–POC relationships with global datasets to test whether separate accounting is needed for petrogenic-carbon-poor systems.

Prior global syntheses linked biogenic POC export to SS yield through exponential regressions (e.g., Galy et al., 2015), estimating 110–230 MtC yr−1 biogenic OC erosion, with most rivers draining petrogenic-carbon-rich sedimentary rocks. Studies highlight event-driven POC fluxes and offshore burial efficiencies exceeding 65% in some high-SS systems (e.g., Bengal Fan, Taiwan). Mountain belts and cyclone-prone regions can dominate erosional POC export. Few assessments exist for humid-tropical igneous islands and island-arc terrains (e.g., Guadeloupe, Hawaii), where petrogenic OC and sulfides are scarce and silicate weathering rates are high, implying strong CO2 drawdown. Previous work in Puerto Rico underestimated POC due to limited high-discharge sampling and reliance on LOI-based POC estimates derived from global averages inappropriate for clay-rich, highly weathered catchments.

Study area: Two small, forested catchments in the Luquillo Mountains, Puerto Rico: Icacos (3.26 km²; granodiorite; rainfall ~5000 mm yr−1; runoff ~3906 mm yr−1) and Mameyes (17.8 km²; volcaniclastic; rainfall ~3720 mm yr−1; runoff ~2770 mm yr−1). Both lack significant petrogenic organic carbon, reduced sulfur, or carbonates. Chemical weathering rates are among the world’s highest for silicates; thick saprolites are present.

Hydrology and sampling: Continuous USGS discharge records were compiled as hourly, gap-free series (1991–2015). >4000 river samples (1991–2015) were collected across hydrographs, with emphasis on large to extreme storms, via manual, depth-integrated, and automated event sampling (ISCO). Suspended and dissolved fractions were filtered (0.2 µm for WEBB; 0.7 µm GF/F in additional campaigns). Extreme-runoff POC samples were collected up to >10 mm hr−1 runoff.

Chemical analyses: POC and particulate nitrogen (PN) were measured on subsets of SS samples via elemental analysis (Costech CHNSO 4010) and coulometry (Coulometrics Model 5120). LOI at 550 °C was measured for many samples but was not used to quantify POC due to high structural water in clays. Given petrogenic-carbon-poor bedrock, all measured river OC is treated as biogenic.

Load and yield modeling: SS datasets were quality-screened (known date/time, concurrent discharge, SS ≥3.5 mg, ≥800 mL filtered from ~1000 mL), excluding ~2.2% of samples; any samples with measured POC/PN were retained regardless. Loads were estimated using USGS LOADEST (model #2; log-transformed), with hourly discharge as the estimation file and measured concentrations as calibration. To extend concentration coverage, POC (mg L−1) was estimated from SS (mg L−1) using site-specific regressions: Icacos POC = 0.054 × SS^0.977 (r²=0.96); Mameyes POC = 0.069 × SS^0.977 (r²=0.91). As a cap, estimated POC was limited to ≤0.5×LOI when 2×POC exceeded measured/estimated LOI. DOC and DIC loads were estimated with LOADEST using WEBB measurements; alkalinity (bedrock-derived, atmospherically corrected) was equated to DIC. Daily LOADEST loads were summed to annual yields; all constituent models had r²>0.93. Load bias for POC was 8.5% (Icacos) and 28.3% (Mameyes).

Rainfall event attribution: A 15-min Icacos rain-gage record (1993–2015) defined events as separated by >3 h without rain. Events were grouped by amount: <22 mm (≤90th percentile; includes dry-period flows), 22–92 mm (90th–98th), and >92 mm (>98th; top 1% extreme events). For Mameyes, a ±6 h buffer around Icacos events was applied (trimmed when overlapping). Daily loads were apportioned to hourly and assigned to events; discharges <50th percentile were excluded.

NPP estimation and NPP_export: Basin-average NPP was computed by area-weighting forest-type NPP using 15 m land-cover maps (1999–2003) and plot-based NPP measurements for Tabonuco, Sierra Palm, Palo Colorado, and elfin forests. Estimated basin NPP: Icacos 990 ± 1 tC km−2 yr−1; Mameyes 1013 ± 85 tC km−2 yr−1. NPP_export (%) = POC yield / NPP × 100. NPP_export for DOC was also calculated.

POC burial and carbon budget: Coastal burial efficiency of terrestrial POC was assumed 22 ± 5% to estimate biogenic POC_burial yield. Geological CO2 sink (tC km−2 yr−1) was computed as POC_burial + DIC, with negligible petrogenic OC or sulfide oxidation sources in these basins.

Global comparisons and regressions: A global dataset of river SS and POC yields was partitioned into petrogenic-carbon-poor (<2% petrogenic POC of total POC) and petrogenic-carbon-rich groups. Log–log regressions were fit for biogenic POC yield vs SS yield and NPP_export (%) vs SS yield, then expressed as power laws for each group. The Luquillo results were compared to prior global regression (Galy et al., 2015) and used to estimate potential underestimation of fluxes from humid-tropical igneous montane islands (~320,000 km²).

• Biogenic POC yields: Icacos 65 ± 16 tC km−2 yr−1; Mameyes 17.7 ± 5.1 tC km−2 yr−1 (higher than prior estimates for these basins).
• SS yields: Icacos 1530 ± 410 t km−2 yr−1; Mameyes 310 ± 96 t km−2 yr−1.
• DOC and DIC yields: Icacos DOC 12.4 ± 1.5; DIC 11.0 ± 0.5 tC km−2 yr−1. Mameyes DOC 8.3 ± 0.9; DIC 13.4 ± 0.6 tC km−2 yr−1.
• Biogenic POC burial (assuming 22 ± 5% efficiency): Icacos 14.3 (+7.7/−6.0) tC km−2 yr−1; Mameyes 3.9 ± ~1 tC km−2 yr−1.
• Geological CO2 sink (POC_burial + DIC): Icacos −25.3 ± ~3 tC km−2 yr−1; Mameyes −17.3 ± 1 tC km−2 yr−1, comparable to the largest known areal sinks (e.g., Whataroa, NZ).
• POC:DOC export ratios: Icacos 5.0 ± 1.0; Mameyes 2.0 ± 0.2, indicating particulate dominance, especially during extreme events.
• NPP_export (%) as POC: Icacos 6.6 ± 1.6%; Mameyes 1.7 ± 0.7%. Icacos exceeded previously reported maxima in ~20–25% of years (annual POC yields up to 97–179 tC km−2 yr−1 in 5 years with multiple storms).
• Event-scale partitioning: Extreme rainfall events (>92 mm; ~1% of events; ~18% rainfall; ~13% runoff) exported 54–63% of annual SS and 52–60% of annual biogenic POC. DIC export (>80%) and about half of DOC export occurred during small events (<22 mm) and dry periods.
• Underprediction by global regression: Applying Galy et al. (2015) SS–POC relation yields 4.7 ± 1.7 (Icacos) and 1.9 ± 0.8 (Mameyes) tC km−2 yr−1, underestimating observed biogenic POC by 92 ± 2% and 87 ± 3%, respectively.
• New regressions distinguishing bedrock groups: • Petrogenic-carbon-poor: Y_bio_pp = 0.100 × Y_SS^0.90 (r²=0.96, n=68); mean biogenic POC% ≈ 9.0 ± 2.8 (n=16). • Petrogenic-carbon-rich: Y_bio_pr = 0.051 × Y_SS^0.64 (r²=0.83, n=84); mean biogenic POC% ≈ 0.87 ± 0.66 (n=84). • NPP_export relations: NPP_export_pp = 0.011 × Y_SS^0.88 (r²=0.96, n=58); NPP_export_pr = 0.010 × Y_SS^0.53 (r²=0.78, n=50).
• Global implication: For humid-tropical igneous montane islands and arcs (~320,000 km²), biogenic POC flux is 0.64–1.57 MtC yr−1 via the Galy regression, but 5.59–23.52 MtC yr−1 using the new petrogenic-carbon-poor regression, implying global biogenic OC export (currently 110–230 MtC yr−1) may be underestimated by ~2.1–8.7% yr−1.

The study demonstrates that forested, humid-tropical, petrogenic-carbon-poor catchments with igneous/volcaniclastic bedrock can be strong geological CO2 sinks, driven by substantial export and presumed burial of biogenic POC together with DIC production from silicate weathering. Crucially, the majority of POC and SS export occurs during rare, extreme rainfall events, underscoring the need to capture high-flow conditions when estimating carbon budgets. The commonly used global SS–POC regressions, derived largely from petrogenic-carbon-rich sedimentary systems, severely underpredict biogenic POC yields in petrogenic-carbon-poor terrains because these systems couple modest SS yields (less erodible bedrock) with high biogenic POC contents and very high rainfall. Distinguishing bedrock types yields improved regressions for both POC yield and NPP_export, aligning predictions with observations from Luquillo and similar settings. The high NPP_export in Icacos suggests that physical erosion currently exceeds soil formation, mobilizing substantial biospheric carbon. Given projected increases in tropical cyclone intensity, event-driven POC exports and thus CO2 sequestration via burial could increase, enhancing the sink role of such catchments. However, the magnitude of the long-term sink hinges on actual marine burial efficiencies near Puerto Rico, which remain unmeasured.

This work provides a 25-year, event-focused quantification of SS and carbon exports from two humid-tropical, petrogenic-carbon-poor rivers in Puerto Rico, revealing exceptionally high biogenic POC yields and strong geological CO2 sinks dominated by extreme rainstorms. It shows that global SS–POC relationships underpredict POC export in such settings and proposes new regressions tailored to bedrock petrogenic-carbon content. Extrapolation suggests a nontrivial upward revision (2–9%) to global biogenic OC export when accounting for humid-tropical igneous mountainous regions. The study underscores the critical role of extreme events in carbon export and the need for their targeted sampling. Future research should: (1) directly measure coastal burial efficiencies adjacent to these rivers, (2) quantify coarse POC fluxes during storms, (3) refine event-based sampling and modeling to reduce load biases, and (4) expand datasets from petrogenic-carbon-poor terrains to further constrain global carbon cycle contributions.

• No direct measurements of biogenic POC burial efficiency in the adjacent marine environment; a global average (22 ± 5%) was assumed.
• Coarse particulate organic carbon was not quantified and may contribute substantially during extreme storms.
• Potential overestimation bias in daily POC loads, particularly for Mameyes (28.3% load bias; modestly above USGS recommended maximum).
• Low-flow POC% estimates are uncertain due to method differences and low SS concentrations; however, annual fluxes are dominated by high flows.
• Some of the largest hurricanes affecting Puerto Rico (e.g., Hugo 1989; Irma and Maria 2017) fall outside the study period (1991–2015), likely leading to conservative long-term export estimates.
• Rainfall event attribution for Mameyes relied on Icacos rain-gage timing with buffering, introducing timing uncertainty.