Environmental DNA signatures distinguish between tsunami and storm deposition in overwash sand

The study addresses the challenge of distinguishing tsunami from storm overwash deposits in the geological record—a key need for reconstructing the frequency and magnitude of past coastal hazards to inform risk assessment and planning. Conventional sedimentological proxies (e.g., grain size, microfossils, geochemistry) can be ambiguous or altered post-deposition, especially in tropical settings. Leveraging microbial responses to pulse disturbances, the authors hypothesize that environmental DNA (eDNA) preserved in sediments contains diagnostic microbial community signatures that can differentiate overwash deposits from intercalating soils/sediments and discriminate tsunami from storm deposits. The research questions are: (i) Do microbial communities differ by location and environment and can individual tsunami and storm deposits be identified from intercalating sediments? (ii) Can modern tsunami and storm deposits be distinguished? (iii) Can the source of overwash deposits be identified from molecular signatures? (iv) Is there a global microbial indicator for overwash deposits?

There is longstanding debate on distinguishing tsunami from storm deposits due to overlapping sedimentary characteristics despite differing depositional mechanisms. Multiproxy methods (grain size, diatoms/foraminifera/ostracods, elemental geochemistry) each have limitations: grain-size can be inconclusive; microfossils may not preserve well in tropics; elemental signatures can be altered by precipitation or microbial activity or have ambiguous sources. Prior microbial studies of tsunami effects often used culture-based methods that capture a small culturable fraction (<5%), limiting interpretability. Metabarcoding studies (e.g., Thailand post-2004 IOT; Japan post-2011 Tohoku) showed community shifts between flooded and unflooded soils and persistence of changes up to years, and identified potential indicators (e.g., sulfur-oxidizers, halotolerant bacteria). However, earlier work rarely examined differences within stratigraphic records at the same location or distinguished tsunami from storm deposits. This study seeks to fill those gaps using high-throughput metabarcoding with robust multivariate statistics.

Study design and sites: Two coastal settings impacted by the 2004 Indian Ocean Tsunami (IOT) and later storms were sampled: (1) Phra Thong Island, Thailand—ridge and swale system with 2004 IOT deposits in organic-rich swales (Swale X, Swale Y) and a 2007 storm overwash behind the berm; (2) Cuddalore, Tamil Nadu, India—sandy beach–dune system with 2004 IOT deposits and 2011 Cyclone Thane deposits preserved in a backdune pit. Permissions were obtained; strict decontamination (20% bleach) and sterile techniques (powder-free gloves) were used to avoid modern DNA contamination. Sampling: Phra Thong: two sediment cores from swales (6 samples from Swale X; 3 from Swale Y), 5 backdune pit samples, beach/intertidal sands, and marine sediments and water from 2 m and 20 m depths. Cuddalore: 30 backdune pit samples spanning aeolian, Cyclone Thane, reworked aeolian, 2004 IOT, and underlying sands; beach and lagoon sediments; marine sediments and waters from 5 m, 9.5 m, and 15 m depths. Samples were transported in liquid nitrogen and stored at −80 °C. Grain-size analysis: Organic matter removed with 15% H2O2; Malvern Mastersizer 3000 measured particle size after 1 min sonication. Statistics (mean phi, sorting) computed using GRADISTAT v9.1 (Folk & Ward, moments methods). Chemical analysis: TOC, TN, TS measured at HKU Stable Isotope Lab. Approximately 30 mg samples acidified (6N HCl), dried (60 °C), combusted and analyzed by EA-IRMS. DNA extraction: From 250 mg sediment using DNeasy PowerSoil (Qiagen) with phenol:chloroform:isoamyl alcohol pre-step, followed by kit protocol; cleanup with Zymo OneStep PCR Inhibitor Removal. Quantified by Qubit. Amplicon library preparation: PCR with two primer sets. For prokaryote-focused universal SSU: 926WF/1392R targeting 16S rRNA V6–V8 across Archaea, Bacteria, and some Eukarya; for eukaryotes: 574F/1132R targeting 18S rRNA V4–V7. KAPA HiFi master mix, standardized thermocycling (20 cycles; 95 °C denaturation; 55 °C annealing; 72 °C extension). Triplicate reactions per sample, pooled and purified (AMPure XP). Sequencing on Illumina MiSeq (2×300 bp, v3 kit) at Macrogen. Sequence processing: Adaptor/primer trimming with cutadapt v2.10; quality filtering and truncation (R1 <280 bp, R2 <230 bp removed; Q≤2 or high expected error reads removed) using DADA2 filterAndTrim. ASV inference with DADA2; taxonomy assigned with SILVA v132 (16S) and PR2 v4.12.0 (18S) using RDP classifier in DADA2. Low-count ASVs (<10 reads) and low-confidence (bootstrap <90% at supergroup/phylum) removed. Final tables: 25,034 ASVs (16S), 5,840 ASVs (18S). Statistics: Analyses in R 3.6.1 with phyloseq and vegan. Read counts normalized to median depth; datasets rarefied to minimum sequences (42,412) for diversity comparisons. Diversity: Shannon (H′) and Simpson (D). Community structure: PCoA (Bray–Curtis); Envfit with TOC, TN, TS (9,999 permutations) to test environmental correlations; additional Envfit with grain-size mean and sorting. Distance-based redundancy analysis (dbRDA) for constrained ordination; ellipses at 80% CI. PERMANOVA (PRIMER v7) on Bray–Curtis similarities to test differences among sample types (five levels); for Phra Thong, two-factor model included year (2014 vs 2015); p-values via 9,999 permutations; dispersions also tested. Differential abundance: DESeq2 (negative binomial GLM; Wald test; adjusted p<0.001) to identify ASVs distinguishing tsunami vs storm deposits; visualization via heatmaps and Ward’s hierarchical clustering. Data and code: Raw sequences at NCBI BioProject PRJNA343068; statistical results and R code at https://github.com/slimelab/Tsunami-microbes.

• Sequencing output: On average 84,011 sequences per sample; total of 25,034 16S/SSU ASVs. Analyses rarefied to 42,412 reads per sample for even depth.
• Diversity patterns: Overall microbial diversity higher at Phra Thong (Shannon H′=6.3283, Simpson D=0.9952) than Cuddalore (H′=5.2393, D=0.9795). Marine sediments showed highest diversity at both sites (Phra Thong: H′=6.6750, D=0.9974; Cuddalore: H′=6.1859, D=0.9961). Overwash deposits had the lowest richness onshore; in Cuddalore Cyclone Thane deposits had higher diversity (H′=5.1262, D=0.9777) than 2004 IOT deposits (H′=4.8647, D=0.9727).
• Community composition: Proteobacteria dominated all samples. Phra Thong hosted diverse Proteobacterial classes (Gamma-, Delta-, Alpha-, Beta-), while Cuddalore was dominated by Gamma- and Alpha-proteobacteria. Archaeal Thaumarchaeota (Nitrososphaerales, Nitrosopumilales) present at both, with lower proportion at Phra Thong. Eukaryotes (18S) dominated by Ciliophora (Colpodea, Spirotrichea) and Cercozoa (Filosa).
• PCoA (Bray–Curtis): Clear separation between marine water and marine sediment communities; separation by location (site effect). When standardized (excluding waters/lagoon), marine/beach/intertidal communities differed from backdune/swale samples (Axis 2 ≈8.1% variance). Overwash deposits were similar to adjacent soils but still distinguishable in constrained analyses.
• Environmental correlates: TOC/TN/TS significantly correlated with site differences (p<0.02) but not with sample-type differences within sites. Grain-size Envfit indicated significant correlations of eDNA patterns with mean or sorting overall (p<0.05), but no correlation between microbial dissimilarity and mean grain size among stratigraphic units (overwash vs intercalated soils).
• dbRDA (Bray–Curtis): Phra Thong—marine/beach/intertidal cluster separate from backdune/swale (CAP1 16% variance; ANOVA F1,15=4.0117, p=0.001). Cuddalore—similar separation (CAP1 25.8%; F1,29=13.4769, p=0.001). Focusing on stratigraphy: Phra Thong—storm deposits separated from tsunami and terrestrial (CAP1 19.2%; F1,11=2.8694, p=0.003); tsunami deposits separated along CAP2 8.6% (F1,11=1.2954, p=0.056). Cuddalore—Cyclone Thane distinct from 2004 IOT and adjacent layers (CAP1 16%; F1,23=4.5135, p=0.001); tsunami marginally separated (CAP2 6.1%; F1,23=1.6988, p=0.081).
• PERMANOVA: Phra Thong—community composition differed among sample types (pseudo-F4,13=2.4865, p<0.0001); pairwise: 2007 storm differed from all others; beach differed from soils/terrestrial. 2004 IOT not significantly different from soils and marine despite very low similarity (<1% to marine; <9% to soils). Cuddalore—all types differed except tsunami vs overlying/underlying aeolian/reworked sediments (pseudo-F4,29=5.674, p<0.0001); higher dispersion in aeolian (>50%) noted. Tsunami <30% similar to storm and <25% to aeolian.
• Temporal stability: Phra Thong sampling in 2014 vs 2015 showed consistent differences among types with no year effect (type×year F2,13=0.8520, p=0.7337; year F1,13=1.0434, p=0.4236).
• Differential ASVs (DESeq2, adj p<0.001): 527 ASVs (Phra Thong) and 92 ASVs (Cuddalore) distinguished storm vs tsunami deposits. Storm deposits (both sites) enriched in Gammaproteobacteria and Actinobacteria. Phra Thong storm-unique families included Blastocatellaceae (Subgroup 4), Subgroup 6, Holophagae, Nitrososphaeraceae, Nitrosomonadaceae, Microscillaceae, Acetobacteraceae. Cuddalore storm-unique families included Parcubacteria, Chromobacteriaceae, Rubinisphaeraceae, Burkholderiaceae, Micromonosporaceae, Bacillaceae, Nocardioidaceae, Sporichthyaceae, Caulobacteraceae, Sericytochromatia. Families Chitinophagaceae and Thermoplasmata were unique to storm deposits at both locations. Cuddalore tsunami deposits contained unique Clostridia and Paenibacillaceae ASVs. No ASVs were uniquely diagnostic of 2004 IOT at Phra Thong.
• Eukaryote metabarcoding (18S): No significant discrimination among storm, tsunami, and terrestrial samples, likely due to primer/amplicon length constraints for diverse eukaryotes.
• Source attribution: No strong similarity between tsunami deposits and offshore marine sediments, suggesting post-disturbance selection favors generalist taxa rather than direct preservation of source-specific marine communities.

The study demonstrates that microbial community signatures preserved in overwash sands can be used to distinguish event deposits from intercalating soils and to discriminate tsunami from storm deposits at two contrasting tropical sites. Site-level differences reflect contrasting geomorphology and sediment geochemistry, but within-site constrained ordinations and PERMANOVA show that storm deposits are consistently distinct, and tsunami deposits are distinguishable though sometimes closer to adjacent aeolian/reworked strata (likely due to mixing during deposition/backflow and post-event reworking). Diversity patterns (lower diversity in overwash vs surrounding sediments) and differential ASV profiles support unique disturbance-associated communities, with recurring storm-associated families (e.g., Chitinophagaceae, Thermoplasmata). Eukaryote community-level metabarcoding was not effective, indicating prokaryote-focused markers are more diagnostic under current protocols. Source tracing to marine vs coastal origins was inconclusive, as tsunami deposits did not closely match marine sediments, consistent with selection for disturbance-tolerant generalists and long-term community shifts after overwash. Overall, eDNA metabarcoding offers a robust, statistically supported tool to aid identification of overwash deposits in the geological record, complementing traditional sedimentology and microfossils.

This work provides the first unambiguous evidence that environmental DNA metabarcoding can discriminate tsunami and storm overwash deposits in sandy coastal settings. Across two Indo-Pacific sites impacted by the 2004 IOT and subsequent storms, prokaryotic community signatures reliably separated event deposits from intercalating sediments and distinguished tsunami from storm deposits. Differential ASV analyses identified suites of taxa characteristic of storm deposits at both sites, although no universal tsunami-specific markers were found. The approach is robust to short-term temporal variability and complements conventional proxies, addressing a major challenge in paleocoastal hazard reconstruction. Future research should (i) evaluate preservation and diagnostic power in older/palaeo-overwash deposits across diverse climates, (ii) refine chemical/metal analyses to identify environmental drivers of microbial shifts, (iii) improve eukaryotic marker strategies with taxon-specific metabarcoding, and (iv) expand site coverage to assess potential global indicator taxa for overwash types. These advances can enhance reconstructions of extreme-wave event histories and improve coastal risk assessments.

• Study focuses on modern (recent) tsunami and storm deposits at two sites; generalizability to older palaeo-deposits and other geomorphic settings remains to be tested. - Eukaryote community-level metabarcoding (18S) did not discriminate deposit types, likely due to primer length and variability constraints with current sequencing read lengths. - Could not conclusively identify environmental factors driving differences between tsunami and storm communities; geochemical variables (TOC, TN, TS) mainly explained between-site, not within-site, differences. - Tsunami deposits at Cuddalore were not strongly distinct from adjacent aeolian/reworked sands, potentially due to mixing during deposition and post-event reworking, and higher dispersion in aeolian samples. - No global, universal microbial marker for tsunami deposits was identified; storm-associated families (e.g., Chitinophagaceae, Thermoplasmata) are promising but require broader validation.