Influence of Dominant Phytoplankton Species on Disinfection By-Product Formation During Active-Substance Ballast Water Treatment: Skeletonema costatum vs. Akashiwo sanguinea

Abstract

Active substance-based Ballast Water Management Systems (BWMS) can generate disinfection by-products (DBPs) by reacting with dissolved organic matter (DOM). However, current IMO G9-based assessments often overlook qualitative DOM variations. This study investigated DBP formation following NaDCC treatment in natural seawater dominated by the diatom Skeletonema costatum and the dinoflagellate Akashiwo sanguinea. Laboratory-cultured DOM was also analyzed using ATR-FT-IR, PCA, and 2D-COS to evaluate structural differences. In field experiments, S. costatum treatment primarily produced brominated trihalomethanes (THMs) and specific haloacetic acids (HAAs) with a limited composition. Conversely, A. sanguinea treatment yielded a diverse range of DBPs, including nitrogenous DBPs (HANs). FT-IR results, supported by 2D-COS, revealed that A. sanguinea-derived DOM underwent non-monotonic structural changes and distinct sequential functional group reactions, suggesting multiple, time-delayed precursor interactions. These findings demonstrate that phytoplankton species-specific DOM composition significantly dictates DBP profiles and temporal dynamics. Therefore, environmental risk assessments for BWMS must incorporate the qualitative characteristics of biogenic DOM and dominant species traits, particularly during coastal bloom events, to ensure more accurate management strategies.

1. Introduction

The transport and spread of invasive aquatic organisms via ballast water is an internationally recognized environmental issue [1,2]. The International Maritime Organization (IMO) adopted the International Convention for the Control and Management of Ships’ Ballast Water and Sediments (Ballast Water Management Convention, BWM Convention) in 2004 to minimize disruption to marine ecosystems caused by ballast water discharge [3].

Among various treatment technologies, the use of active substance-based BWMS is widely applied due to their high disinfection efficiency and the advantage of maintaining residual disinfectant potency within the ballast water tank [4]. However, during the treatment process using active substances, disinfection by-products (DBPs) can be generated through reactions with dissolved organic matter (DOM) in seawater [5]. Trihalomethanes (THMs), haloacetic acids (HAAs), haloacetonitriles (HANs), and some other nitrogen-containing DBPs have been reported to possess potential toxicity to aquatic organisms [6]. Particularly in marine environments, the presence of bromide ions can increase the potential for the formation of brominated DBPs [5].

Reflecting these concerns, BWMS using active substances undergo the IMO approval process, which comprises multiple assessment components (e.g., efficacy and safety evaluations). In particular, an environmental risk assessment of ballast water discharge is required under the G9 procedure, and final approval is contingent upon the assessment results [7]. Under the current testing and evaluation framework, salinity conditions and DOM concentration are treated as key variables influencing DBP formation. In land-based tests, the organic matter level in test water is typically controlled by adding substitute organic substances such as glucose, lignin, and cellulose derivatives [8]. However, this approach assumes that DBP formation potential is primarily governed by the “quantity (concentration)” of DOM. It may also fail to adequately reflect the compositional diversity of biogenic DOM, which dominates in natural coastal and harbor waters—that is, the “quality” of DOM [4,9,10].

The composition of DOM in natural aquatic environments varies substantially depending on the relative contributions of allochthonous (land-derived) and autochthonous (marine biogenic) organic matter. Particularly in coastal and harbor waters, marine biogenic DOM becomes the primary source during phytoplankton blooms. DOM is generated and accumulated through active and passive efflux of phytoplankton photosynthetic products, cell lysis due to aging and death, sloppy feeding during microbial and zooplankton consumption, and enzymatic hydrolysis of particulate organic matter by bacteria [11,12,13]. DOM derived from phytoplankton enters seawater through extracellular exudation, metabolites, and cell breakdown processes. Compared to humic substances of terrestrial origin, it may have lower aromatic content and relatively prominent nitrogen-containing functional groups (e.g., proteinaceous and amino acid components) [14,15]. The biochemical composition of this phytoplankton-derived DOM can vary depending on the taxonomic group and species of the producing organisms. Differences in the composition ratio between nitrogen-containing organic matter (e.g., proteins, amino acids) and carbohydrate-based components alter the characteristics of the primary precursors for DBPs during active substance treatment processes. Consequently, not only the pathways for DBP formation but also their temporal behavior, including generation, transformation, and persistence, may vary depending on the species composition of the phytoplankton providing the DOM [4,16,17].

Nevertheless, most studies on DBPs under BWMS treatment conditions have focused on water quality parameters such as DOM and Specific UV Absorbance (SUVA) [4,18,19], species-comparative experimental studies examining how structural differences in DOM originating from different phytoplankton species influence the initial composition of DBPs formed during active substance treatment and the subsequent transformation and compositional changes in DBPs over reaction time remain limited [4,17]. As many previous studies have been conducted based on freshwater and municipal wastewater treatment processes, there remains a gap in research simultaneously considering the presence of bromine in marine environments and coastal bloom conditions (where biogenic DOM predominates).

Fourier transform infrared spectroscopy (FT-IR) is an analytical technique reflecting major functional groups and polymeric characteristics in complex mixed organic samples, and has been utilized to compare the overall structural characteristics of DOM [20,21,22]. Although FT-IR has limitations in precisely identifying individual molecules, it can compare and interpret trends in DOM composition at the functional group level. This is achieved by interpreting spectra repeatedly acquired under identical conditions using multivariate analysis (PCA, PLS, etc.) or two-dimensional correlation spectroscopy (2D-COS) based on external perturbation variables such as time [23,24,25].

In this study, we focus on species-dependent differences in disinfection by-product (DBP) formation during active substance-based ballast water treatment in seawater, with particular emphasis on how structural characteristics of phytoplankton-derived DOM influence DBP composition and temporal behavior.

2. Materials and Methods

2.1. Research Design Overview

This study conducted two independent experimental components—field-based treatments using naturally phytoplankton-dominant seawater and laboratory-based culture experiments—to compare DBP formation characteristics and relate them to species-specific DOM structural features derived from Skeletonema costatum (Greville) and Akashiwo sanguinea (Gert Hansen & Moestrup).

(1) Field-based DBP Formation Characterization: Natural seawater samples were collected during periods when S. costatum (thereafter SC conditions) or A. sanguinea (thereafter AS conditions) were dominant in actual marine environments. These samples contained DOM characteristics formed by each dominant species under field conditions and were used to evaluate DBP formation characteristics reflecting real marine conditions.

(2) Species-Specific Structural Analysis of DOM Derived from Laboratory-Cultured Phytoplanton: To more clearly compare the structural differences in species-derived DOM, independent laboratory-based cultivation experiments were conducted. Since natural seawater contains mixed organic matter from various sources, making it difficult to isolate species-specific contributions, S. costatum (hereafter Skel-cult) and A. sanguinea (hereafter Aka-cult) were each cultured as single species. The recovered cells were rinsed with 0.2 µm filtered seawater to remove residual culture medium components, resuspended in sterile filtered seawater, treated under the same active substance conditions as the field experiment, and then subjected to FT-IR analysis.

The two experiments were designed to independently characterize DBP formation under field conditions and the structural characteristics of species derived DOM, with the results subsequently compared and integrated during interpretation. Water quality parameters & DBP data obtained from the TRO 10 treatment of A. sanguinea included in this study were previously reported in Cha et al. (2024) [26].

2.2. Natural Seawater Collection and Phytoplankton Composition (DBP Experiment)

Samples were collected from the KIOST research pier located in Jangmok-bay, Geoje, Republic of Korea, during March 2016 when S. costatum was dominant and November 2016 when A. sanguinea was dominant. In situ water temperature and salinity were measured using a multiparameter water quality sonde (YSI 6600, YSI Inc., Yellow Springs, OH, USA), and surface seawater was collected using a Niskin bottle (DAIHAN Co., Incheon, Republic of Korea).

Phytoplankton cell density was determined by microscopic counts using a Sedgewick–Rafter counting chamber (1 mL). Cells were counted in fields (or the entire chamber) and converted to cell density (cells mL⁻¹) based on the chamber volume. During the period when S. costatum was dominant, the total biological density of natural seawater samples was approximately 50,000 cells mL⁻¹, with S. costatum accounting for over 99% of the total population. In contrast, the total phytoplankton biomass density of natural seawater sampled during the period dominated by A. sanguinea was 3010 cells mL⁻¹, with A. sanguinea accounting for 2160 cells mL⁻¹ or over 72% of the total population. The initial pH of the collected seawater, measured prior to active substance addition, was 8.14 (SC condition) and 8.10 (AS condition), determined using a benchtop pH meter (Orion Star A211, Thermo Scientific, Waltham, MA, USA).

For both conditions, two concentration ranges-high biomass (H) and low biomass (L) -were established in the natural seawater-based test water to evaluate the effects of biomass concentration. Biomass concentration was adjusted by diluting the natural seawater with filtered seawater, and 20 L of test water was prepared for each concentration range.

The biomass concentrations for each condition are as follows.

• S. costatum: High biomass 50,000 cells mL⁻¹ (SC-H), Low biomass 1000 cells mL⁻¹ (SC-L)
• A. sanguinea: High biomass 2160 cells mL⁻¹ (AS-H), Low biomass 420 cells mL⁻¹ (AS-L)

2.3. Active Substance Injection and Reaction Conditions (DBP Experiment)

The active substance used was sodium dichloroisocyanurate (NaDCC; Wako Pure Chemical Ind., Osaka, Japan). A specific amount of NaDCC was thoroughly dissolved in a 50 mL conical tube and then added to each biological concentration interval. The target total residual oxidant (TRO) level was set at 10 mg L⁻¹ for all test tubes. The treated water was stored at 20 °C under dark conditions. Samples for DBP and water quality parameter analysis were collected 24 h (D1) and 120 h (D5) after treatment [27], while an untreated seawater sample (B0) was simultaneously collected on the sampling day and analyzed in parallel to characterize background DBP and water quality parameter levels and in the field seawater.

2.4. Disinfection By-Product (DBP) Analysis

Samples B0, D1 and D5 were immediately neutralized with sodium thiosulfate solution to remove residual oxidants (TRO < 0.1 mg/L) before analysis for DBPs. Referencing the candidate substance groups proposed in the IMO G9 methodology, this study selected and quantified THMs, HAAs, and HANs. For sample preservation, acid was added to THMs to achieve pH < 2, while HAAs and HANs were preserved by adding ammonium chloride (10 mg) and 1–2 drops of 6 M hydrochloric acid. All samples were collected in amber glass bottles, refrigerated at 4 °C or below, and analyzed within 14 days. THMs were analyzed by GC/MS (7890B GC–5977A MSD, Agilent Inc., Santa Clara, CA, USA) according to US EPA Method 8041A. HAAs and HANs were analyzed by GC/ECD (7890A GC, Agilent Inc., USA) according to US EPA Methods 555.2 and 551.1, respectively.

2.5. Total Residual Oxidants (TRO) and Dissolved Organic Carbon/Organic Nitrogen (DOC/DON) Analysis

TRO was measured immediately after oxidant injection and mixing (0 h), and at D1 and D5, using a portable residual chlorine meter (Pocket Colorimeter II, Hach Inc., Loveland, CO, USA) employing the DPD (diethyl-p-phenylene diamine) method.

Dissolved Organic Nitrogen (DON) concentration was calculated as the difference between Total Dissolved Nitrogen (TDN) concentration and Dissolved Inorganic Nitrogen (DIN) concentration. Samples for TDN and DIN concentration measurement were pretreated by filtration through GF/F filters (pore size 0.7 µm) and analyzed using a nutrient automatic analyzer (QuAAtro 39, Seal Analytical Ltd., Wrexham, UK).

For the analysis of dissolved organic carbon (DOC), samples were gravity-filtered through GF/F filters (pre-combusted at 450 °C) to remove particulate matter. The filtrate was acidified with 50% H3PO4 to achieve pH ≤ 2. Subsequently, DOC concentrations were determined using a Total Organic Carbon analyzer (TOC-V_CPH, Shimadzu Inc., Tokyo, Japan) via the high-temperature catalytic combustion method.

All samples were stored frozen at −20 °C or below after pretreatment until analysis. For quality assurance/quality control (QA/QC), a method blank and a calibration check standard were included every 10 samples. Water-quality parameters were measured in triplicate as analytical replicates (repeated measurements of the same prepared sample) to assess analytical precision. Variability is reported as mean ± SD of analytical triplicates.

2.6. Carbon-Equivalent Biomass Estimation and Carbon-Normalized DBP Yields

To facilitate cross-condition comparisons, DBP concentrations were additionally expressed as carbon-normalized yields (µg mg C⁻¹). Carbon-equivalent biomass (mg C L⁻¹) was estimated from microscopic cell abundance and measured cell biovolume. Cell biovolume (V, µm³) was calculated from measured cell dimensions using appropriate geometric approximations. Cellular carbon content (C, pg C cell⁻¹) was then estimated from V using published carbon–volume relationships: for diatoms, C = 0.288 × V^0.811 for dinoflagellates, C = 0.760 × V^0.819 [28]. Carbon-equivalent biomass concentration (mg C L⁻¹) was calculated by multiplying C by cell abundance (cells mL⁻¹) and converting units (Table S1). Carbon-normalized DBP yield was calculated as:

$$DBPs yields \left(\mu g mg C^{-1}\right)={\displaystyle \frac{\left[DBP\right](\mu g L^{-1})}{\left[C\right](mg C L^{-1})}}$$

2.7. Phytoplankton Cultivation and Cell Washing for FT-IR Analysis (Species-Derived DOM)

FT-IR analysis was performed to compare the relative structural characteristics of DOM among species, using laboratory-cultured cells for both S. costatum and A. sanguinea. Sterile filtered seawater was prepared by filtering and sterilizing seawater from Jangmokman Bay at approximately 32 psu salinity. Cultures were maintained at 20 °C under 12L:12D conditions in medium supplemented with F/2 medium.

To minimize the influence of background DOM present in the culture medium itself, cultured cells were filtered through a 20 µm nylon sieve to completely remove the existing culture medium. Cells retained on the sieve were gently rinsed with 0.2 µm filtered seawater. The recovered cells were immediately resuspended in sterile filtered seawater, adjusting the final cell density to 50,000 cells mL⁻¹ for S. costatum and 2000 cells mL⁻¹ for A. sanguinea.

2.8. FT-IR Measurement Conditions, Sample Collection and Spectral Preprocessing for Multivariate·2D-COS Analysis

Each reconstituted sample (2 L) was injected with NaDCC to adjust the total residual oxidant (TRO) concentration to 10 mg L⁻¹, then reacted under dark conditions at 20 °C. Samples were collected at 1 day (D1) to 5 days (D5) post-treatment, with the sample immediately prior to treatment used as the background condition (B0).

The collected samples and the background sample were filtered through GF/F filters calcined at 450 °C to remove particulate matter, and approximately 30 mL of filtrate was recovered. The filtrate was stored frozen at −20 °C or below, then freeze-dried. The dried DOM samples were pulverized and homogenized using an agate mortar. FT-IR analysis was performed using an ATR-FT-IR spectrometer (PerkinElmer Inc., Hopkinton, MA, USA). The acquired spectra underwent baseline correction and normalization preprocessing using PerkinElmer Spectrum IR software (ver. 10.7.2).

After processing, the difference spectrum was derived for each spectrum (Dt) acquired at each reaction time by calculating ‘Δ Absorbance = Absorbance (Dt) − Absorbance (B0)’. The calculated difference spectrum was used to compare and interpret the relative increase or decrease patterns (peak enhancement/reduction, band shape changes) over time in the major functional group regions, rather than for the identification of individual compounds. Furthermore, principal component analysis (PCA) and two-dimensional correlation spectroscopy (2D-COS) were performed to summarize the overall variation characteristics of the FT-IR spectra. PCA was computed in a Python (version 3.12.0) environment using NumPy, Pandas, and scikit-learn libraries, with the spectral data mean-centered prior to analysis.

The 2D-COS analysis calculated synchronous and asynchronous correlation spectra based on the time perturbation data. The synchronous correlation spectrum was computed as the covariance-based matrix product of the dynamic spectral matrix, while the asynchronous correlation spectrum was calculated by applying the Hilbert-Noda transform. Correlation spectra were visualized as contour plots using Matplotlib (version 3.10.3). Reaction sequence analysis utilized the sign combinations of synchronous and asynchronous cross-peaks according to the Noda rule [29].

Since the asynchronous correlation spectrum may contain minute cross-peaks due to numerical noise, an absolute value-based threshold condition was applied to ensure the reliability of the analysis [23]. In this study, only cases where the absolute magnitude of the asynchronous correlation value exceeded ε were considered valid sequential changes. ε was set to a value clearly distinguishable from the noise level in the asynchronous correlation spectrum (ε = 0.005).

3. Results

3.1. Changes in TRO (Total Residual Oxidants), DOM (Dissolved Organic Matter) and DIN (Dissolved Inorganic Nitrogen) Following Active Substance Treatment

Changes in TRO and DOM-related indicators (DOC and DON) during the active substance treatment process were compared using natural seawater collected during the period when S. costatum and A. sanguinea were dominant (SC and AS conditions). Since untreated samples were not obtained under identical time conditions, this study utilized them as background conditions rather than controls to evaluate relative changes due to treatment.

Following active substance injection under SC conditions, TRO concentrations showed a gradual decrease over time (

Table 1. Changes in nutrient concentrations, total residual oxidant (TRO), dissolved organic carbon (DOC), dissolved inorganic nitrogen (DIN), dissolved organic nitrogen (DON), Salinity, and pH under Skeletonema costatum and Akashiwo sanguinea-dominated conditions following active substance treatment.

Parameter	DIN	DON	TRO	DOC	NH4+	Salinity	pH
Unit	μM	μM	mg/L	mg/L	μM	psu
1SC-2B-3D0	6.96 ± 0.19	1.88 ± 0.14	4NA	2.36 ± 0.03	3.99 ± 0.02	31.05	8.14
SC-5H-D1	22.8 ± 0.59	25.1 ± 0.15	6.60 ± 0.70	5.01 ± 0.10	0.87 ± 0.01		No data
SC-H-D5	25.5 ± 0.44	24.1 ± 0.16	3.40 ± 0.04	3.93 ± 0.02	0.31 ± 0.03
SC-L-D1	1.57 ± 0.43	10.0 ± 0.07	6.85 ± 0.11	4.83 ± 0.04	0.06 ± 0.01
SC-L-D5	1.78 ± 0.07	11.6 ± 0.01	6.10 ± 0.12	4.41 ± 0.14	6 <MDL
AS-B-C0	10.0 ± 0.14	10.8 ± 0.40	NA	5.42 ± 0.07	4.02 ± 0.04	31.30	8.10
AS-H-D1	4.56 ± 0.04	30.8 ± 0.05	0.90 ± 0.11	12.4 ± 0.10	2.39 ± 0.01		No data
AS-H-D5	5.02 ± 0.15	21.8 ± 0.30	<MDL	9.17 ± 0.11	2.44 ± 0.02
AS-L-D1	25.9 ± 0.30	18.7 ± 0.21	5.50 ± 0.10	5.22 ± 0.01	3.36 ± 0.01
AS-L-D5	22.9 ± 0.32	15.3 ± 0.44	2.70 ± 0.15	7.59 ± 0.13	<MDL

Note: ¹SC and AS denote field seawater collected during Skeletonema costatum and Akashiwo sanguinea dominance, respectively; ² untreated seawaters collected on the sampling day; ³D0, D1, and D5 represent days 0 (pre-treatment), 1, and 5 after active substance treatment. ⁴NA: not applicable; ⁵L and H indicate low and high biomass conditions; ⁶MDL: Method Detection Limit. Water quality parameter data obtained from the TRO 10 treatment of A. sanguinea presented in this table were previously reported in Cha et al. (2024) [26].

). After TRO 10 mg L⁻¹ treatment, residual oxidants were detected in both SC-L and SC-H samples until D1 and D5, with a relatively larger decrease observed in SC-H. The TRO concentration at D1 was 6.60 mg L⁻¹ in SC-H and 6.85 mg L⁻¹ in SC-L, decreasing to 3.40 mg L⁻¹ and 6.10 mg L⁻¹ at D5, respectively. In contrast, TRO concentrations under the AS condition showed a markedly faster reduction over time (Table 1). After the initial 10 mg L⁻¹ treatment, residual oxidants were significantly lower in both AS-L and AS-H compared to the SC condition, with the most rapid decrease observed in the high-biomass group (AS-H). Specifically, the TRO concentration in AS-H was 0.90 mg L⁻¹ at D1 and decreased to below the detection limit (0.00 mg L⁻¹) by D5. In AS-L, the TRO concentration was 5.50 mg L⁻¹ at D1 and declined substantially to 2.70 mg L⁻¹ by D5. Overall, the AS condition showed a faster TRO decay than the SC condition.

DOC showed an increasing trend after active substance treatment in both SC and AS conditions (Table 1). For SC conditions, the DOC concentration in the pre-treatment background sample (SC-B) was 2.36 ± 0.03 mg L⁻¹. By D1, it increased to 5.01 ± 0.10 mg L⁻¹ (SC-H) and 4.83 ± 0.04 mg L⁻¹ (SC-L). Although it decreased slightly by D5, it remained above background levels. The DOC increase was greater in the high-biomass group (H) than in the low-biomass group (L). Under AS conditions, DOC concentrations also increased significantly at D1 compared to the background sample (AS-B), reaching a maximum of 12.4 ± 0.10 mg L⁻¹. Although they decreased by D5, they remained above background levels.

DON also showed an increasing trend after active substance treatment, but its variability was relatively greater than that of DOC. Under SC conditions, DON showed a particularly large increase in SC-H. Under AS conditions, a decreasing trend was observed from D1 to D5. Differences were observed between SC and AS conditions in the characteristics of TRO reduction and the patterns of DOC and DON changes. Under AS conditions, TRO decreased rapidly alongside high DOC and DON concentrations. In contrast, under SC conditions, despite increased DOM concentrations, residual TRO was maintained for a relatively long period.

Changes in dissolved inorganic nitrogen (DIN) and ammonium (NH₄⁺) concentrations differed between SC and AS conditions following active substance treatment (Table 1). Under SC conditions, DIN increased substantially under SC-H treatments, rising from 6.96 μM at D0 to 22.8–25.5 μM at D1 and D5, whereas under SC-L, DIN remained relatively low (1.57–1.78 μM). In contrast, NH₄⁺ concentrations under SC conditions decreased sharply after treatment, declining from 3.99 μM at D0 to ≤0.87 μM at D1 and reaching near depletion by D5 (0.31 μM or <MDL).

Under AS-dominated conditions, DIN exhibited different temporal behavior depending on biomass level. In AS-H, DIN decreased relative to the initial level (10.0 μM) and stabilized at approximately 4.56–5.02 μM after treatment. Conversely, AS-L showed elevated DIN concentrations at D1 (25.9 μM) followed by a slight decrease by D5 (22.9 μM). NH₄⁺ concentrations under AS conditions were comparatively higher than those under SC treatments, remaining between 2.39 and 3.36 μM at D1 before decreasing to <MDL at D5 in low-biomass conditions.

3.2. Formation and Concentration Characteristics of DBPs Under Different Phytoplankton-Dominated Conditions

The concentration and temporal variation patterns of disinfection by-products (DBPs) generated after active-substance treatment differed between SC and AS conditions (

Table 2. Concentration (μg/L) of disinfection by-products (DBPs) grouped by chemical class under Skeletonema costatum and Akashiwo sanguinea dominance at different biomass levels and treatment times.

Compound (Abbreviation)	1SC-2B0	SC-3H-4D1	SC-H-D5	SC-L-D1	SC-L-D5	AS-B0	1AS-H-D1	AS-H-D5	AS-L-D1	AS-L-D5
	Trihalomethanes (THMs), μg/L
Dichloromethane (DCM)	5ND	ND	ND	ND	ND	ND	2.91	0.00	2.20	0.00
Dichlorobromomethane (DCBM)	ND	ND	ND	ND	ND	ND	2.02	2.24	0.00	2.09
Dibromochloromethane (DBCM)	ND	6.70	10.89	6.06	6.56	ND	13.82	14.71	3.96	4.24
Bromoform	ND	179.1	302.7	141.3	198.0	ND	286.8	290.1	95.1	109.0
Σ THMs	ND	185.8	313.6	147.4	204.6	ND	305.6	307.1	101.3	115.3
	Haloacetic acids (HAAs), μg/L
Monochloroacetic acid (MCAA)	ND	ND	ND	ND	ND	ND	0.00	2.72	1.03	0.76
Monobromoacetic acid (MBAA)	ND	3.59	7.53	3.36	4.61	ND	1.47	8.27	1.40	1.24
Trichloroacetic acid (TCAA)	ND	ND	ND	ND	ND	3.19	0.45	0.28	0.17	0.17
Bromochloroacetic acid (BCAA)	ND	1.80	5.38	1.42	2.722	1.25	4.65	10.8	4.77	5.28
Dibromochloroacetic acid (DBCA)	ND	19.6	18.9	12.7	5.72	0.83	0.47	41.5	0.37	0.89
Dibromoacetic acid (DBAA)	2.18	41.9	100.8	35.9	50.6	1.61	62.3	107.3	39.9	5.97
Dichlorobromoacetic acid (DCBA)	ND	2.99	8.14	2.37	2.72	ND	1.86	7.95	1.05	1.53
Tribromoacetic acid (TBAA)	1.01	31.0	64.9	26.2	30.9	ND	11.22	48.7	7.49	8.52
Σ HAAs	3.19	286.8	519.3	229.4	301.9	6.88	82.4	227.6	56.2	24.4
	Haloacetonitriles (HANs), μg/L
Monochloroacetonitrile (MCAN)	ND	ND	ND	ND	ND	ND	0.33	1.16	0.00	0.00
Dichloroacetonitrile (DCAN)	ND	ND	ND	ND	ND	ND	0.00	0.53	1.03	0.00
Monobromoacetonitrile (MBAN)	ND	ND	ND	ND	ND	ND	1.53	9.24	0.23	0.00
Bromochloroacetonitrile (BCAN)	ND	0.38	0.34	0.41	0.24	ND	0.00	0.95	0.48	0.53
Dibromoacetonitrile (DBAN)	ND	7.60	5.33	9.51	1.57	ND	ND	ND	ND	ND
Σ HANs	ND	7.98	5.67	9.92	1.81	ND	1.86	11.87	1.74	0.53
ΣDBPs	3.19	294.7	524.9	239.3	303.7	6.88	389.9	546.5	159.2	140.2

Note: ¹SC and AS denote field seawater collected during Skeletonema costatum and Akashiwo sanguinea dominance, respectively; ²B0: untreated seawaters collected on the sampling day; ³L and H indicate low and high biomass conditions; ⁴D1 and D5 represent days 1 and 5 after active substance treatment; ⁵ND: Not detected (below the detection limit). DBP data obtained from the TRO 10 treatment of A. sanguinea presented in this table were previously reported in Cha et al. (2024) [26].

, Figure 1 and Figure 2). This section reports absolute concentrations (µg L⁻¹) of individual DBPs and DBP groups (THMs, HAAs, and HANs) at D1 and D5 and additionally presents carbon-normalized DBP yields (µg mg C⁻¹), as shown in Figure 1c,d.

DBPs were also measured in the untreated seawaters collected on the sampling day (B0) to characterize background levels. In SC-B0, THMs and HANs were not detected (ND), whereas HAAs were detected at a low level (ΣHAAs = 3.19 µg L⁻¹), primarily as DBAA (2.18 µg L⁻¹) and TBAA (1.01 µg L⁻¹). In AS-B0, THMs and HANs were also ND, but HAAs were higher than in SC-B0 (ΣHAAs = 6.88 µg L⁻¹), with TCAA (3.19 µg L⁻¹), DBAA (1.61 µg L⁻¹), BCAA (1.25 µg L⁻¹), and DBCA (0.83 µg L⁻¹) detected (Table 2).

In the NaDCC-treated samples (D1 and D5), DBP concentration increased relative to B0. Under SC conditions, THMs with a high degree of bromine substitution were predominantly formed following active substance treatment, with bromoform detected at the highest concentration under all conditions (Table 2). In SC-H, the bromoform concentration increased from 179.1 μg L⁻¹ at D1 to 302.7 μg L⁻¹ at D5. In SC-L, it also maintained a relatively high concentration, rising from 141.3 μg L⁻¹ at D1 to 198.0 μg L⁻¹ at D5. Dibromochloromethane (DBCM) was detected in the range of 6.70–10.89 μg L⁻¹ under SC-H and 6.06–6.56 μg L⁻¹ under SC-L.

Among HAAs, dibromoacetic acid (DBAA) and tribromoacetic acid (TBAA) were identified as the major DBPs (Table 2). Under SC-H, DBAA increased significantly from 41.9 μg L⁻¹ at time point D1 to 100.8 μg L⁻¹ at time point D5, while TBAA also increased from 31.0 μg L⁻¹ to 64.0 μg L⁻¹. Under SC-L, DBAA increased from 35.9 μg L⁻¹ to 50.6 μg L⁻¹, and TBAA increased from 26.2 μg L⁻¹ to 30.9 μg L⁻¹, but the increase was relatively smaller compared to the high-concentration conditions. For HANs, dibromoacetonitrile (DBAN) was characteristically detected under SC conditions. Under SC-H, it remained stable at 7.60 μg L⁻¹ at D1 and 5.33 μg L⁻¹ at D5. Under SC-L, it decreased from 9.51 μg L⁻¹ at D1 to 1.57 μg L⁻¹ at D5. Conversely, monobromoacetonitrile (MBAN) and other HANs were not detected under SC conditions.

Under AS conditions, a wider variety of DBP species were detected compared to SC conditions, and the concentration distribution of DBPs changed significantly with treatment time (Table 2). Among THMs, bromoform was identified as a major DBP even under AS conditions, though its concentration range was lower than under SC conditions. Under AS-H, the bromoform concentration at D1 was 286.8 μg L⁻¹, and it remained at 290.1 μg L⁻¹ at D5. Conversely, dichloromethane (DCM) and dichlorobromomethane (DCBM) showed distinct concentration variations among THMs, being detected only under certain conditions or not detected at the D5 time point. HAAs showed the most pronounced concentration increase over time in the DBP group under AS conditions. DBAA increased from 62.3 μg L⁻¹ at D1 to 107.3 μg L⁻¹ at D5 under AS-H, and TBAA also increased significantly from 11.2 μg L⁻¹ to 48.7 μg L⁻¹ under the same conditions. DBCA increased to 41.5 μg L⁻¹ at D5 under AS-H, indicating a tendency for the absolute concentrations of HAAs to be concentrated in the later stages under AS conditions (Table 2). For nitrogen-based DBPs (HANs), the test water from A. sanguinea showed a more diverse and distinct formation pattern than under SC conditions, both in terms of types and concentrations. Monobromoacetonitrile (MBAN) increased significantly from 1.53 μg L⁻¹ at D1 to 9.24 μg L⁻¹ at D5 under AS-H, while dichloroacetonitrile (DCAN) and bromochloroacetonitrile (BCAN) were also detected under some conditions. This increase in HAN concentrations was not observed or was limited under SC conditions.

Group-summed DBP concentrations are summarized in Figure 1a,b. Under SC conditions, THMs increased from 185.8 to 313.6 µg L⁻¹ in SC-H and from 147.4 to 204.6 µg L⁻¹ in SC-L between D1 and D5, while HAAs increased from 100.9 to 205.7 µg L⁻¹ in SC-H and from 82.0 to 97.3 µg L⁻¹ in SC-L. HANs decreased from 7.98 to 5.67 µg L⁻¹ in SC-H and from 9.92 to 1.81 µg L⁻¹ in SC-L. Under AS conditions, THMs were 305.6–307.1 µg L⁻¹ in AS-H and 101.3–115.3 µg L⁻¹ in AS-L; HAAs increased from 82.4 to 227.6 µg L⁻¹ in AS-H but decreased from 56.2 to 24.4 µg L⁻¹ in AS-L; and HANs increased from 1.86 to 11.9 µg L⁻¹ in AS-H while decreasing from 1.74 to 0.53 µg L⁻¹ in AS-L (Figure 1a,b).

The corresponding carbon-normalized DBP yields are shown in Figure 1c,d. Under SC conditions, THM yields were 18.8–31.8 µg mg C⁻¹ in SC-H and 746.6–1036 µg mg C⁻¹ in SC-L, and HAA yields were 10.2–20.8 µg mg C⁻¹ in SC-H and 415.1–493.0 µg mg C⁻¹ in SC-L. HAN yields were 0.57–0.81 µg mg C⁻¹ in SC-H and 9.19–50.2 µg mg C⁻¹ in SC-L (Figure 1c, Table S2). Under AS conditions, THM yields were 25.7–25.8 µg mg C⁻¹ in AS-H and 43.8–49.8 µg mg C⁻¹ in AS-L; HAA yields were 6.93–19.1 µg mg C⁻¹ in AS-H and 10.5–24.3 µg mg C⁻¹ in AS-L; and HAN yields were 0.16–1.00 µg mg C⁻¹ in AS-H and 0.23–0.75 µg mg C⁻¹ in AS-L (Figure 1d, Table S2).

3.3. Changes in the FT-IR Spectrum of DOM and Temporal Patterns

Difference spectra (Dt − D0) were calculated by subtracting the pre-treatment spectrum (D0) from the post-treatment spectra (D1-D5) in the normalized FT-IR spectra. The results showed that changes were concentrated in the carbonyl/amide region (1800–1500 cm⁻¹) and the fingerprint region (1200–900 cm⁻¹) (Figure 3a,c). In contrast, while spectral intensity fluctuations were observed in the X-H stretching region (3700–2800 cm⁻¹), no consistent pattern enabling the derivation of a time-dependent relationship was observed.

In Skel-cult, similar wavelength changes in the difference spectra were repeatedly observed from the initial treatment stage (D1) through the intermediate and late stages (D4–D5), indicating a tendency for DOM structural changes to be relatively sustained throughout the reaction period (Figure 3b). In contrast, Aka-cult exhibited non-monotonic changes: the magnitude of variation in the difference spectra peaked at the initial treatment stage (D1), then decreased or shifted in spectral pattern as the reaction progressed (Figure 3d).

PCA analysis targeting the 1800–900 cm⁻¹ range, which contains functional group information directly linked to DOM composition in both cultured phytoplankton species, revealed that the first principal component (PC1) explained the majority of the total variance for both species (Skel-cult: ~79%, Aka-cult: ~72%) (Figure 4). In the PCA score plot, Skel−cult showed a PC1 score moving in a consistent direction after D1, with some reversal observed at later time points; however, the dispersion width of the movement path was relatively limited (Figure 4a). In contrast, Aka-cult exhibited a significant shift in PC1 score at D1 compared to pre-treatment (D0), followed by a tendency to move in the opposite direction at subsequent time points, which was repeatedly observed (Figure 4b). PCA loading analysis confirmed that the wavelength range corresponding to the fingerprint region contributed most significantly to PC1 separation in both species, with particularly high loading values observed around 1150–1115 cm⁻¹. Conversely, the wavelength range of the carbonyl/amide region tended to contribute to the lower principal components (Figure 4c,d).

For time-perturbation-based analysis, 2D-FTIR-COS was performed. In the synchronous map, the strongest autopeak was consistently observed in the fingerprint region (approximately 1160–1100 cm⁻¹) for both Skel-cult and Aka-cult (Figure 5). This indicates that this wavelength region is the most sensitive to active substance treatment, consistent with the PCA loading results. Differences between species were observed in the synchronous cross-peak distribution: Skel-cult showed relatively diverse cross-peak distribution between the fingerprint region and the carbonyl/amide region, whereas Aka-cult exhibited prominent correlation signals within the fingerprint region’s wavelength range (Figure 5a,c). Asynchronous map analysis revealed that cross-peaks meeting the criterion (|Ψ| ≥ ε) were either limited or absent in the carbonyl/amide region. Conversely, in the fingerprint region, numerous asynchronous cross-peaks met the criterion in both species, providing sufficient signals to derive time-sequence-based relationship changes (Figure 5b,d).

Based on the PCA loading patterns and the 2D-FTIR-COS autopeak characteristics, representative bands for Noda-rule analysis were selected to capture changes in both the carbonyl/amide region (amide I, ~1650 cm⁻¹; amide II, ~1540 cm⁻¹) and the fingerprint region (fingerprint-1, ~1100 cm⁻¹; fingerprint-2, ~1030 cm⁻¹). For each selected wavelength pair, the signs of synchronous (Φ) and asynchronous (Ψ) cross-peaks were extracted (

Table 3. Summarizes the temporal relationships between major functional groups in dissolved organic matter (DOM) inferred from synchronous (Φ) and asynchronous (Ψ) Two-dimensional correlation spectroscopy (2D-COS) cross-peaks. The combined signs of Φ and Ψ indicate whether two spectral bands change in the same or opposite directions and which band responds earlier during active substance treatment.

1Species	2Band 1	Band 2	3ν1 (cm−1)	ν2 (cm−1)	4Φ (Synchronous)	Ψ (Asynchronous)
Skel-cult	Amide II	Amide I	1540	1650	5+	−
Skel-cult	Fingerprint-1	Amide I	1100	1650	−	+
Skel-cult	Fingerprint-1	Amide II	1100	1540	−	−
Skel-cult	Fingerprint-2	Amide I	1030	1650	+	−
Skel-cult	Fingerprint-2	Amide II	1030	1540	+	−
Skel-cult	Fingerprint-2	Fingerprint-1	1030	1100	−	+
Aka-cult	Amide II	Amide I	1540	1650	+	−
Aka-cult	Fingerprint-1	Amide I	1100	1650	+	+
Aka-cult	Fingerprint-1	Amide II	1100	1540	−	+
Aka-cult	Fingerprint-2	Amide I	1030	1650	+	+
Aka-cult	Fingerprint-2	Amide II	1030	1540	−	+
Aka-cult	Fingerprint-2	Fingerprint-1	1030	1100	+	+

Footnote: ¹Skel-cult, cultured Skeletonema costatum; Aka-cult, cultured Akashiwo sanguinea. ² Band assignments were defined as Amide I (~1650 cm⁻¹), Amide II (~1540 cm⁻¹), Fingerprint-1 (~1100 cm⁻¹), and Fingerprint-2 (~1030 cm⁻¹). ³ν₁ and ν₂ correspond to Band 1 and Band 2. ⁴Φ and Ψ denote the signs of synchronous and asynchronous cross-peaks obtained from two-dimensional correlation spectroscopy (2D-COS), respectively. ⁵The sign of Φ indicates whether the two bands vary in the same direction (Φ > 0) or in opposite directions (Φ < 0) under the applied perturbation, whereas the sign of Ψ, used together with Φ, provides the relative sequential order of changes according to Noda’s rule. The sequential order is determined from the sign combination as follows: Positive (+)/negative (−) signs represent the direction and temporal relationship of spectral changes, not the magnitude of variation.

) and interpreted using the Noda rule to derive the relative order of spectral changes between functional-group-related bands.

For Skel-cult, the Noda-rule interpretation indicated that amide I (~1650 cm⁻¹) generally responded earlier relative to several fingerprint-region bands, although the inferred order involving amide II (~1540 cm⁻¹) depended on the specific fingerprint band considered (Table 3). Within the fingerprint region, fingerprint bands (~1100 cm⁻¹) tended to precede fingerprint bands (~1030 cm⁻¹) (Table 3). For Aka-cult, sequential relationships were dominated by fingerprint-region dynamics: significant Noda pairs consistently indicated that fingerprint-region bands changed earlier than amide I (~1650 cm⁻¹), whereas amide II (~1540 cm⁻¹) frequently preceded fingerprint-region bands in the significant pairs (Table 3).

4. Discussion

4.1. Species-Specific Differences in DOM Structural Changes Confirmed via FT-IR Analysis

In this study, changes in DOM structure over time following active substance treatment showed species-specific variations as revealed by difference spectra, PCA, and 2D-FTIR-COS analysis of FT-IR spectra [29,30]. Subtracting the spectra after treatment (D1–D5) from the normalized FT-IR spectra before treatment (D0), changes were concentrated in the carbonyl/amide region (1800–1500 cm⁻¹) and the fingerprint region (1200–900 cm⁻¹) for both phytoplankton species. However, the temporal patterns of change and relative contributions were clearly distinguishable between species [24,30]. For Aka-cult, a non-monotonic pattern was repeatedly observed: the magnitude of spectral changes expanded significantly at the initial treatment stage (D1), followed by either a decrease or a subsequent increase in changes at later time points. This suggests that DOM does not undergo a simple cumulative oxidation or decomposition process, but rather forms a dynamic reaction system where a specific precursor pool reacts rapidly early in treatment, followed by subsequent reactions that reorganize the DOM composition [31,32]. In contrast, Skel-cult exhibited repeated changes not only initially but also at later time points (D4–D5), though the directionality of these changes tended to be relatively more consistent than in Aka-cult. This difference was further clarified in the PCA results [33].

4.2. Complexity of the DOM Precursor Pool Suggested by PCA Results

PCA performed on the 1800–900 cm⁻¹ range, which contains functional group information directly related to DOM composition, showed that PC1 explained the majority of the total variance in both species (Skel-cult: approximately 79%, Aka-cult: approximately 72%). This indicates that changes in DOM structure following active substance treatment are dominated by relatively few major functional group alterations [24,33]. However, distinct differences were observed in the temporal trajectory of the PC1 score between the two species. For Skel-cult, the PC1 score moved in a consistent direction after the initial treatment (D1), showing some reversal after D4. Overall, the dispersion width and directional changes in the trajectory were limited. These results suggest that DOM derived from Skel-cult possesses a structurally relatively homogeneous precursor pool and follows a relatively simple reaction pathway even during the active substance treatment process [34].

In contrast, Aka-cult exhibited a pattern where the PC1 score shifted sharply at D1 compared to pre-treatment (D0), followed by a repeated movement in the opposite direction between D3 and D5. This back-and-forth movement can be interpreted as reflecting a process where different precursor pools sequentially react or transform, rather than DOM structural changes proceeding along a single reaction pathway [31].

PCA loading analysis results also supported these species-specific differences. In both species, the fingerprint region (particularly around 1150–1115 cm⁻¹) contributed most significantly to PC1 separation. However, while Skel−cult showed a more dominant contribution from this region, Aka-cult exhibited supplementary contributions from the carbonyl/amide wavelength range. This suggests that multiple functional group pools simultaneously drove changes in the PC space [30,35].

4.3. 2D-FTIR-COS and Noda Rule Results for Time Sequence Comparison

To more precisely analyze the sequential relationship of DOM structural changes due to temporal perturbations, 2D-FTIR-COS analysis was performed [29]. In the synchronous map, the strongest autopeak was observed around 1118–1114 cm⁻¹ in the fingerprint region for both Skel-cult and Aka-cult, indicating this wavelength range is the most sensitive to treatment with the active substance. These results are consistent with the findings from PCA loading analysis that the fingerprint region is a major variance-contributing interval [30].

Application of the asynchronous map and Noda rule revealed quantitative differences in the temporal sequence between the two phytoplankton. For Skel−cult, the sequence derived showed amide I (1650 cm⁻¹) responding earlier for both fingerprint-1 (around 1100 cm⁻¹) and fingerprint-2 (around 1030 cm⁻¹). Furthermore, amide II (1540 cm⁻¹) showed opposite reaction orders for fingerprint-1 and fingerprint-2, suggesting that the inferred order involving amide II depends on the specific fingerprint band considered (Table 3). These results suggest that in Skel-cult-derived DOM, proteinaceous or amide-type functional groups may react preferentially or possess structures more closely associated with the fingerprint series [34].

In contrast, Aka-cult showed a different timing pattern. Noda pairs satisfying the criterion (|Ψ| ≥ ε) indicated that fingerprint-region bands (~1100 cm⁻¹ and ~1030 cm⁻¹) responded earlier than amide I (~1650 cm⁻¹), whereas amide II (~1540 cm⁻¹) often preceded fingerprint-region bands in the significant amide–fingerprint pairs (Table 3). These results are consistent with temporally decoupled responses between carbohydrate/polysaccharide-related and nitrogen-containing functional-group pools in Aka-cult derived DOM [35]

Thus, the Noda rule-based 2D-FTIR-COS results suggest that the temporal priority of DOM structural changes differs between the two phytoplankton culture species, A. sanguinea and S. costatum. Specifically, in Aka-cult, the significant Noda pairs are consistent with more heterogeneous (potentially independent or parallel) responses among functional-group-related bands, whereas in Skel-cult, the inferred sequence indicates a comparatively more coherent and integrated progression of changes [31].

4.4. Relationship Between FT-IR Results and DBP Formation Characteristics

DBP analysis results show that, relative to background levels measured in the untreated waters (B0), DBP concentrations increased. Under AS conditions, THMs, HAAs, and HANs are all observed, with the dominant DBP class changing over treatment time [6,10]. Specifically, under AS-H, THMs formed at high concentrations already by D1, whereas the relative contribution of HAAs and HANs significantly increased by D5. This indicates that DBP formation does not terminate with a single precursor reaction but involves different precursor pools reacting sequentially over time [9,17].

This DBP diversity can be qualitatively and quantitatively linked to FT-IR analysis results. First, the non-monotonic changes in PCA scores and fingerprint/amide band intensities in Aka-cult suggest that carbohydrate-based precursors and nitrogen-containing precursors react or convert at different time points during the active substance treatment process [34,35]. This suggests that carbohydrate-based reactions initially dominate, contributing to THM formation, while later, nitrogen-containing precursors participate in the reaction, providing structural conditions that can lead to HANs and some HAAs [16,36]

Second, the significantly higher levels of brominated DBPs (Br-DBPs) observed under AS conditions compared to SC conditions indicate that the bromination pathway was more active under seawater conditions [13,37]. The strong reorganization and repetitive changes in the fingerprint region observed in FT-IR support the possibility that a reaction environment was formed where polysaccharides and low-molecular-weight organic acid series precursors, sensitive to bromination reactions, were continuously generated and consumed [32,34]. This structural diversity ultimately enables the coexistence of brominated THMs and HAAs, increasing the diversity of DBP composition.

Third, the increase in HANs at the later time point (D5) under AS conditions logically connects the delayed reaction characteristics of amide functional groups observed in Noda rule results [16,29]. That is, nitrogen-containing precursors initially react relatively slowly, but after carbohydrate-based precursors are depleted, they participate in the reaction pathway, leading to a shift in DBP classes over time [17,35]. In contrast, Skel-cult showed a dominant fingerprint region in FT-IR analysis but limited temporal separation from the amide group. This aligns with DBP formation results where THMs and HAAs predominate, while HANs contribute relatively little [6,10].

In this study, FT-IR analysis played a key role not as a tool for directly predicting DBP concentrations, but rather in elucidating how the DOM precursor pool is structurally reorganized during active substance treatment and which functional group pool reacts earlier or later in time [29,30]. Specifically, combining PCA with 2D-FTIR-COS enabled quantitative comparison of the primary driving domains and temporal sequence of DOM structural changes. However, environmental factors not directly observable by FT-IR, such as halide ion concentration, TRO depletion characteristics, pH, and nitrogen form (DON, NH₄⁺), also contribute to the final determination of DBP diversity [15,38]. Therefore, the DBP diversity observed under AS conditions should be understood as a result of the precursor structural complexity confirmed by FT-IR combined with these environmental factors [23,39].

4.5. Interpretation of Differences in DOM Characteristics and DBP Formation Between Diatoms and Dinoflagellates

This study compared the FT-IR spectral changes and DBP production characteristics of dissolved organic matter (DOM) derived from diatoms (S. costatum) and dinoflagellates (A. sanguinea) under active substance treatment conditions. Previous studies have reported that these two biological taxonomic groups exhibit distinct trends in fatty acid composition (EPA vs. DHA), elemental ratios, and biochemical characteristics due to differences in evolutionary background, cell structure, and physiological strategies [40,41,42]. These taxonomic differences may also influence the composition and reactivity of organic matter produced and released by phytoplankton. The observed FT-IR spectral changes across different regions in this study can be interpreted within this context [34,43].

FT-IR analysis revealed that DOM derived from the two phytoplankton species exhibited distinct temporal changes in the amide, carbohydrate, and fingerprint regions. This suggests that the relative contributions and structural transformation processes of protein- and carbohydrate-based organic matter proceeded differently between species, and these differences can be interpreted as being associated with variations in DBP production characteristics after treatment with active substances [35,36]. In particular, differences in fatty acid composition and the proportion of nitrogen-containing organic matter may act as factors influencing the qualitative composition of the DBP precursor pool, consistent with existing reports on the biochemical characteristics of different taxonomic groups [12,40].

However, it has been noted that the composition and reactivity of phytoplankton-derived DOM can vary significantly not only based on taxonomic group-specific characteristics but also depending on physiological state, growth stage, and environmental conditions [12,44,45]. Therefore, the differences observed in FT-IR spectra and DBP production characteristics between diatoms and dinoflagellates in this study should be interpreted as results specific to the experimental conditions and dominant species combinations [25,46]. In natural coastal waters, the occurrence and dominance patterns of diatoms and dinoflagellates are also reported to vary significantly depending on seasonality and environmental factors [47,48]. This variability must be considered when interpreting DOM characteristics of phytoplankton origin.

These differences in cell volume and structure can directly influence DOM release rates, precursor composition, and reactivity during active substance processing [12,49]. Particularly, even under identical biomass density conditions, areas dominated by dinoflagellates may exhibit a relatively higher contribution from nitrogen-containing organic matter and complex precursors. This could act as a factor altering the pathways and temporal transition patterns of DBP formation [12]. Therefore, the results observed in this study suggest that even when the same BWMS and identical operating conditions are applied in waters or ports dominated by different phytoplankton taxa, the composition and potential environmental hazards of the generated DBPs may differ.

5. Conclusions

This study suggests that the characteristics of DBPs generated during active substance-based ballast water treatment processes can vary qualitatively depending on the dominant phytoplankton taxon, beyond simple dissolved organic matter concentration or biomass levels. Experiments using natural seawater collected during periods when the diatom S. costatum and the dinoflagellate A. sanguinea were dominant in actual marine waters revealed distinct differences in the concentration levels of individual DBPs and their transition patterns over time between the two conditions.

Under S. costatum-dominated conditions, highly brominated THMs and HAAs were formed as major DBPs from the early stages of treatment and remained relatively stable throughout the treatment period. In contrast, under A. sanguinea-dominated conditions, a wider variety of DBPs were generated during treatment, including HAAs and nitrogen-based DBPs (HANs), compared to the S. costatum-dominated conditions. These differences were associated with variations in phytoplankton species-derived DOM structural characteristics and temporal response sequences identified through FT-IR-based analysis. Notably, the complex and non-monotonic structural changes observed in DOM derived from dinoflagellates are associated with diversity of DBP formation pathways and the potential for temporal transitions.

Furthermore, the results suggest that even when the same BWMS and identical operating conditions are applied, the composition and potential environmental hazards of generated DBPs may differ across marine areas or ports with different dominant phytoplankton taxa. This raises the possibility that the current BWMS environmental risk assessment framework, which primarily relies on organic matter concentration or standardized surrogate organic matter conditions, may not sufficiently reflect the biological characteristics of actual coastal and port waters (particularly DOM characteristics driven by species during phytoplankton bloom conditions).

In summary, this study indicates that DOM composition differences based on phytoplankton taxonomic groups can influence the formation characteristics and temporal behavior of DBPs during active substance treatment processes. It proposes the need for a more sophisticated approach, considering dominant species characteristics by sea area and season, when establishing environmental hazard assessment and management strategies for BWMS discharge water in the future.