This study focuses on the land–ocean interface of the Mackenzie River Delta–southern Beaufort Sea, a fluvial-dominated delta at the terminus of the Mackenzie River, which drains the largest watershed in Canada (1.8 million km²). Approximately 42 % of the basin is covered by continuous and discontinuous permafrost (Woo and Thorne, 2003; Holmes et al., 2012; Elshamy et al., 2025) (Fig. 1a). This broad permafrost transition exerts strong controls on hydrology, sediment delivery, and organic carbon composition entering the delta-shelf system. Zones of continuous permafrost are characterized by flashy, surface-dominated runoff with strong seasonal pulses and episodic sediment delivery. The shallow soils in continuous permafrost zones can also be characterized by more labile carbon inputs from recent plant material in the near-surface area. As permafrost becomes patchier in discontinuous and sporadic zones, there is more surface-subsurface connectivity, which can lead to older, more processed inputs of carbon (Vonk et al., 2019).

The Mackenzie River is characterized by strong seasonality in both discharge and biogeochemistry (Fig. 2). Discharge observations used for this study were collected from the Mackenzie-Red River gauge station at Tsiigehtchic, part of the Arctic Great Rivers Observatory (ArcticGRO) long-term monitoring effort (arcticgreatrivers.org). The hydrography of the Mackenzie River can be described in three periods: spring ice break-up, summer to autumn months, and winter. For the purpose of this study, we will focus on the ice-free summer and autumn months (June through October).

The in-situ data presented in this study are synthesized from several datasets from both ship and land-based expeditions to the Mackenzie River Delta and Beaufort Sea that took place from 2009 to 2024: MALINA 2009 (Massicotte et al., 2021), Nunataryuk 2019 (Lizotte et al., 2023), PeCaBeau 2021 (Bröder et al., 2022), and FLOCHAR 2024 (this study), in addition to two stationary long-term sampling efforts, ArcticGRO and Mackenzie Monitoring (mackenzie-monitoring.awi.de). Table 1 shows a summary of sampling periods, sample/data types and the measured parameters of individual campaign datasets.

Sampling locations from these datasets include large parts of the Beaufort Shelf, shallow coastal waters around the Mackenzie Delta including Kugmallit Bay and Shallow Bay, and channels and distributaries of the Mackenzie River (Fig. 1b). These data were collected at various levels of river discharge (Fig. 2), where 2024 exhibited drought conditions (purple hydrograph) and 2021 exhibited a high flow year (red hydrograph). The variations exhibited by the Mackenzie River over the sampling periods allow us to investigate the influence of discharge on the transport and transformation of carbon from land to sea.

During ship-based expeditions aboard the CCGS Amundsen (MALINA, PeCaBeau), water samples were taken using Niskin bottles on a Rosette system at defined depths along predefined transects across the shelf. For land-based expeditions (Nunataryuk, FLOCHAR, Mackenzie Monitoring), a combination of small draft boats, hovering helicopter, and riverbank sampling was used to sample freshwater locations within the delta and the shallow receiving waters of the two main outflow regions of the Mackenzie River: Kugmallit Bay to the east, and Shallow Bay and Mackenzie Bay to the west.

During land-based expeditions (Nunataryuk, FLOCHAR), discrete water samples were collected as close to the surface as possible using a pump and plastic hose to fill 10 L containers. For the Mackenzie Monitoring program, river water was sampled at the shore just below the water surface with 1 L bottles. The pump and sample containers were rinsed twice with sample water prior to collection. All water samples were filtered within 12 h of collection (see details on filtration below).

Since this study focuses on characterizing bio-optical relationships for satellite retrievals, only surface water samples (<5 m water depth; assuming a homogenous freshwater upper layer) were included in the compiled dataset. Details of the aforementioned sampling techniques can be found in the following literature for each campaign (Massicotte et al., 2021; Bröder et al., 2022; Lizotte et al., 2023; Morgenstern and Gottschalk, 2023).

The following section describes the methods used to collect and analyze biogeochemical parameters presented in this study. The methods for each parameter across individual campaigns and monitoring projects are slightly different. Differences significant to this study are explained below, but for information beyond what is described, further methodological details on these campaigns are provided in Massicotte et al. (2021), Bröder et al. (2022), Lizotte et al. (2023), and Morgenstern and Gottschalk (2025).

For DOC, the MALINA and Nunataryuk, samples were filtered through 25 mm, Millipore Whatman GF/F pre-combusted filters with a nominal pore size of 0.7 µm. For the FLOCHAR and Mackenzie Monitoring programs, 47 mm, 0.45 µm Millipore cellulose acetate (CA) filters were used. The filtration setup varied slightly between each campaign. For the MALINA campaign, water samples were gravity filtered from the Niskin bottles used for collection. For the Nunataryuk, PeCaBeau, and

Mackenzie Monitoring, 30 mL syringe filtration was used. A Nalgene filtration unit was used for FLOCHAR samples. All filtered samples were acidified and stored frozen (-20 °C) immediately after processing in pre-combusted borosilicate glass vials. DOC concentrations were measured using a high-temperature catalytic oxidation (TOC-VCPH, Shimadzu) at the Alfred Wegener Institute (AWI, Potsdam, Germany) (Benner and Strom, 1993). For ArcticGRO, samples were analyzed on a Shimadzu TOC/TN analyzer at the Woodwell Climate Research Center (WHRC) in the United States.

Water samples for the analysis of colored dissolved organic matter absorption (aCDOM) were filtered and measured using slightly different procedures across the individual campaigns. For the Mackenzie Monitoring and FLOCHAR programs, samples were filtered through 0.45 µm cellulose acetate (CA) syringe filters and subsequently analyzed at the German Research Center for Geosciences (GFZ, Potsdam, Germany) using a PerkinElmer LAMBDA 950 UV–Vis spectrophotometer. Sample processing and analytical protocols followed Juhls et al. (2021). Samples collected during the PeCaBeau, MALINA, and Nunataryuk campaigns were filtered through 0.2 µm GHP syringe filters (Acrodisc Inc.) and analyzed with a Liquid Waveguide UltraPath system (World Precision Instruments Inc.) immediately after sampling aboard ship (MALINA and PeCaBeau) or following short transport to the laboratory (Nunataryuk). For ArcticGRO, absorbance was measured between 200–800 nm at 1 nm intervals using a Shimadzu UV-1800 spectrophotometer and a 1 cm pathlength.

Samples for POC from Nunataryuk and PeCaBeau were filtered using pre-combusted 47 mm Millipore Whatman GF/F filters, while samples from FLOCHAR, Mackenzie Monitoring, and ArcticGRO used pre-combusted 25 mm Millipore Whatman GF/F filters. For Nunataryuk, 0.25–1.6 L of sample volume was filtered, while 2–10 L of sample volume was filtered for PeCaBeau, and 5–6 different volumes of sample were filtered for each station ranging from 0.05–0.4 L for the FLOCHAR campaign. All sample filters were dried at room temperature, 50 °C, or freeze-dried. For Nunataryuk, samples were weighed three times each with a Mettler Toledo microscale at Laval University (Quebec City, Canada). Two punched replicates from the filters were placed into tin capsules and processed in a PerkinElmer elemental analyzer (PE 2400 Series-II CHNS/O Elemental Analyzer). For Mackenzie Monitoring, POC content was measured using a punched portion of the GF/F filters, fumigated and dried while analysis was completed on an Elemental Analyzer (EA) interface coupled to a stable isotope analyzer (Thermo Scientific EA IsoLink IRMS System) using standards Ethylenediaminetetraacetic acid (EDTA, Sigma Aldrich), following Lattaud et al. (2025). For FLOCHAR, POC concentration was processed at Helmholtz-Zentrum Hereon (Geesthacht, Germany), where the samples were acidified to remove carbonates, pressed in tin capsules, and the particulate carbon concentration determined in a Vario TOC cube instrument (Elementar, Germany). For PeCaBeau, the sample filters were not washed for salts after filtration, which may underestimate the resulting measurement, but we do not believe this has a significant impact.

For particulate absorption (aP) samples from Nunataryuk, approximately 20–500 mL of water was filtered through 25 mm Whatman GF/F pre-combusted filters. A spectrophotometer (Cary 100; Agilent Technologies, Inc.) equipped with a small (60 mm) integrating sphere was used to measure the transmittance and reflectance from the sample filters (350 to 800 nm at 1 nm increments) (Tassan and Ferrari, 1995, 2002). For PeCaBeau, water samples were filtered through pre-combusted 25 mm Whatman GF/F filters. The filters were stored frozen until measurements of particulate absorption were made at Scripps Institution of Oceanography (San Diego, California, USA) using a custom-made spectrophotometer attached to a 150 mm integrating sphere. For FLOCHAR, one volume (3–30 mL) was filtered on a combusted 25 mm GF/F filter for samples measured in the visible (350–800 nm), and one volume (100–400 mL) was filtered on a combusted 25 mm GF/F filter for samples measured in the near infrared (NIR, 700–2500 nm). All sample filters were directly measured with a QFT-ICAM (Röttgers et al., 2015). Here, measurements with a PSICAM (Röttgers and Doerffer, 2007) conducted on unfiltered and filtered water samples (500 mL; 0.2 µm) are used for accurate aP measurements in the wavelength range 400–700 nm, and finally combined with the results of the QFT-ICAM.

For SPM samples from Nunataryuk and PeCaBeau, pre-combusted 47 mm Millipore Whatman GF/F filters were used, while pre-combusted 25 mm Millipore Whatman GF/F filters were used for FLOCHAR and Mackenzie Monitoring. For Nunataryuk, samples were weighed three times each with a Mettler Toledo microscale. For FLOCHAR, 5–6 different volumes of the same sample (0.1–2 L) were filtered using a Nalgene filtration unit as described in Röttgers et al. (2014). For Nunataryuk and Mackenzie Monitoring, 30 mL syringe filtration was used. For all samples, values for SPM were obtained by subtracting the initial weight of the blank filters from the final weight of the dried particulate-matter-laden filter. For PeCaBeau, the sample filters were not washed for salts after filtration, which may overestimate the resulting measurement, but we do not believe this has a significant impact.

Inherent optical properties (IOPs) such as light absorption, beam attenuation, and scattering coefficients define apparent optical properties (AOPs), including Rrs, which is often used synonymously with ocean color. To relate water sample properties to radiometric measurements, absorption at 443 nm was analyzed alongside DOC and POC concentrations from co-located samples collected during all campaigns. This is a commonly used wavelength in ocean color remote sensing due to its sensitivity to multiple optically active constituents and its widespread availability across satellite missions (Pegau et al., 1997; Sullivan et al., 2006). Identifying and characterizing these variabilities is necessary to understand and define the limits of validity of these bio-optical relationships in order to best estimate DOC and POC across the fluvial-marine compartments. In order to evaluate the limitations of these relationships, we compared the relationship of each compartment (river, delta, coastal, marine) with a cross-compartment relationship that spans the full fluvial-marine and salinity gradient.

For this study, filtration for particle absorption in the visible (350–800 nm) called for 2–20 mL of sample filtered through a combusted 25 mm GF/F filter. These filters were directly measured with a quantitative filter technique using an integrating cavity absorption meter (QFT-ICAM) as described in Röttgers et al. (2015). Because this study synthesizes datasets collected during several independent campaigns between 2009 and 2024, minor methodological differences in sample collection, filtration, and analytical protocols exist among the contributing datasets. In particular, different filter types and pore sizes were used for dissolved measurements (0.7 µm GF/F filters during MALINA and Nunataryuk while 0.45 µm cellulose acetate filters were used during FLOCHAR and the Mackenzie Monitoring program). Additionally, CDOM absorption was measured using different optical systems (e.g., UltraPath/Liquid Waveguide systems versus conventional spectrophotometers). Such differences can potentially influence measured DOC concentrations and CDOM absorption by altering the operational definition of the dissolved fraction or by introducing small instrument-specific biases.

However, previous methodological intercomparisons indicate that these effects are generally small relative to the natural variability observed across Arctic fluvial marine gradients. For example, filter comparison experiments in the Lena River showed that DOC concentrations measured with 0.45 and 0.7 µm filters differed on average by only ∼ 2.5 %, while aCDOM(443) measurements obtained using different filtration approaches and analytical instruments typically varied by ∼ 5 %–10 % without showing a systematic bias (Supplement Fig. S1 in Juhls et al., 2021). These differences are small compared with the large spatial gradients observed in this study.

While such methodological variability may introduce minor scatter in cross-campaign comparisons and bio-optical regressions, it is unlikely to affect the major patterns or relationships identified across the fluvial-marine continuum. Nonetheless, the compilation of long-term multi-campaign datasets highlights the importance of continued efforts toward methodological harmonization in Arctic monitoring programs to further improve the comparability of optical and biogeochemical measurements.

Radiometric measurements were conducted using a variety of instrument setups for both ship and land-based expeditions to capture the in-situ water reflectance spectra (Rrs). Table 2 provides detailed information on sampling techniques for each sampling campaign.

In-situ measurements of bio-optical properties are essential for the calibration and validation of OCRS algorithms. When integrated with radiometric observations (e.g., upwelling radiance, downwelling irradiance, and water-leaving reflectance), these measurements provide critical ground-truth data to improve satellite-derived estimates of POC and DOC concentrations. For further interpretation of these bio-optical and radiometric measurements, we employed the OWT classification scheme by Bi and Hieronymi (2024).

This method distinguishes clusters of Rrs, which are determined by different concentration ranges of the optically active water constituents. The classification scheme is designed to apply to inland, coastal and ocean waters, as well as for various hyper- and multispectral band settings of in-situ measurement methods and satellite missions. In this method, three optical variables are extracted from the Rrs spectrum to classify the OWT, leveraging both the spectral shape and magnitude. These are: the apparent visible wavelength (AVW) between 400 and 800 nm, the (Box-Cox-transformed) spectral area below three red, green, and blue bands (ABC), and the normalized difference index at green and red (NDI). Ten types are defined, covering the optical extremes from oligotrophic ocean waters (OWT 1), to eutrophic waters with high phytoplankton biomass (OWT 5), and to CDOM-rich waters (OWT 7). OWT labels ending with “a” and “b” are intended for types with similar spectral shapes but different magnitudes, with “b” generally representing a variant with higher brightness. In color scales, OWT 1 and 2 stand for blue, 3 for turquoise, 4 and 5 for greenish, 6 for light brown and 7 for dark brown waters. For the purposes of this study and to better represent the regional optical variability, broader categories river, delta, coastal, and marine are used as well.

To examine the fate and variability of biogeochemical parameters across the fluvial-marine transition of the Mackenzie River Delta-Beaufort Sea system, we generated a large dataset that covers spring freshet to late summer for years between 2009 and 2024 (Table 1, Figs. 1 and 2). Compared to previously published studies, this synthesis presents a unique compilation of riverine, deltaic distributary, coastal and marine water types, which we use to examine biogeochemical relationships and calculate flux of organic matter across the land-sea transition.

The compiled dataset was first classified into three water types according to in-situ salinity measurements (CastAway CTD) based on the Venice System (1958): freshwater (salinity < 0.5 psu), coastal (0.5–20.0 psu), and marine (>20 psu). To further examine the variability within the dynamic delta zone of the Mackenzie River Delta, we subclassified freshwater based on channel morphology: main river channel (river compartment) upstream of the distributary network channels (delta compartment) (Fig. 1b). The river compartment includes the main stem of the Mackenzie River channel upstream of its bifurcation into distributary channels, located at Tsigethchiic, which is approximately 150 km from the river mouth (Tarasenko et al., 2023). The delta compartment includes distributary channels downstream of the first major bifurcation at the delta apex, where interaction between the river and deltaic channels continue downstream until the coastal waters.

The four aquatic compartments classified by salinity and channel morphology are used in this study to describe the spatial patterns of DOC, POC, and SPM across the fluvial-marine transition. The main riverine channel is hereafter referred to as the “river” compartment, freshwater distributary channels of the deltaic floodplain as the “delta” compartment, waters with salinity 0.5–20.0 psu as the “coastal” compartment, and waters >20.0 psu as the “marine” compartment.

Sampling coverage was not evenly distributed across seasons or compartments per campaign, reflecting the multi-campaign nature of the compiled dataset. The distribution of DOC, POC, and SPM concentrations was examined using all available data throughout the campaign years (2009–2024) and the two monitoring stations. DOC concentrations were highest in the river compartment (n=39; mean 4.78 ± 1.12 mg L⁻¹), and declined steadily across the fluvial-marine gradient towards marine waters (n=50; mean 1.88 ± 0.96 mg L⁻¹; Fig. 3a). POC exhibited a broadly similar decrease along the gradient, but showed greater variability within the delta compartment (n=91; mean 1.38 ± 1.21 mg L⁻¹) compared to the river compartment (n=45; mean 1.49 ± 1.06 mg L⁻¹).

Similarly, SPM concentrations were highest in the river compartment (n=39; mean 104.63 ± 86.20 mg L⁻¹) and decreased considerably towards marine waters (n=56; mean 7.17 ± 12.44 mg L⁻¹; Fig. 3), with the largest variability observed in river and delta waters. Sample availability varied among compartments and seasons, reflecting differences in campaign focus and accessibility across the fluvial–marine transition.

Examining the seasonality of the river compartment, where year-round observations are available from the ArcticGRO monitoring station at Tsiigehtchic, SPM and POC concentrations show peaks during the freshet (May–June) and remain elevated into summer (August–September). A second, smaller and more variable peak occurs during the late summer/fall (September–October) (Fig. 3c and d). DOC concentrations exhibit a similar summer maximum, followed by increased concentrations during autumn and early winter (October–December) (Fig. 3e).

Concentrations of dissolved and particulate constituents notably decreased with increasing salinity along the fluvial-marine transition (Fig. 4). DOC exhibited a strong negative linear relationship with salinity (0.2–35 psu) (r2=0.79), with the highest concentrations (>10 mg L⁻¹) occurring at very low salinities and declining to <1 mg L⁻¹ at salinities above 30 psu (Fig. 4a).

Figure 4b depicts aCDOM(443) showing a similar inverse relationship with salinity (r2=0.74), with values decreasing from >4 m⁻¹ in freshwater to much lower values in marine waters. POC and SPM concentrations also generally declined with increasing salinity, although with greater scatter relative to dissolved components (r2=0.5 and r2=0.45, respectively). POC concentrations were highest at low salinities (>3 mg L⁻¹) and decreased to much lower concentrations in marine waters (Fig. 4). Similarly, SPM concentrations were highest in the freshwater endmember (up to 400 mg L⁻¹) and declined substantially along the salinity gradient, with most marine samples exhibiting concentrations below 20 mg L⁻¹ (Fig. 4d).

The dashed red lines show the theoretical conservative mixing line between freshwater (≤0.2 psu) during spring freshet and the marine endmember. The dashed black lines show the theoretical conservative mixing line of year-round averaged concentrations between freshwater (≤0.2 psu) and the marine endmember. Red-outlined symbols represent water samples collected during the spring freshet (May through June) across different campaigns. The spring freshet mean for DOC (6.61 mg L⁻¹) is noticeably higher than the annual freshwater mean (5.12 mg L⁻¹) (Fig. 4a). Similarly, for aCDOM(443), the spring freshet measurements draw the freshwater mean from (1.45 m⁻¹ annual mean) to (2.01 m⁻¹ spring freshet mean) (Fig. 4b). Particulates exhibit a similar pattern between year-round and spring freshet freshwater endmembers. For SPM, the average freshwater freshet concentration is almost 2× higher than the annual freshwater average (117.03 and 62.97 mg L⁻¹, respectively) (Fig. 4d), while the average freshwater freshet concentration for POC (1.62 mg L⁻¹) is also noticeably higher than the mean annual freshwater endmember (1.02 mg L⁻¹) (Fig. 4c).

Overall, these relationships demonstrate pronounced decreases in both dissolved and particulate organic matter pools along the fluvial–marine mixing continuum with noticeable influences from the seasonal spring freshet from the Mackenzie River.

Across the salinity gradient, total OC concentrations declined from freshwater to marine environments (Fig. 5a). This decrease was primarily driven by a reduction in the dissolved fraction of OC, which constituted the predominant fraction throughout the fluvial-marine gradient, while POC contributed a smaller and more variable fraction. The most pronounced reduction in total OC occurred at low salinities (<10 psu), followed by a more gradual decline towards marine waters. Variability was highest at low salinity, where both DOC and POC showed a broader range of concentrations, whereas offshore waters showed lower concentrations with reduced variability. When aggregated by compartment (Fig. 5b), the river and delta compartments exhibited the highest total OC concentrations, with DOC consistently exceeding POC, while the marine compartment showed the lowest total OC and the smallest relative POC contribution to the OC pool. Note that sample numbers varied among salinity bins and compartments, reflecting uneven spatial and seasonal sampling coverage across the fluvial marine transition.

Comparison of particulate and dissolved absorption from filtered and unfiltered water samples revealed clear differences along the fluvial-marine continuum. Particulate absorption (aP) consistently dominated over the dissolved component (aCDOM) along the salinity gradient (Fig. 5c), with a distinct decrease in aP above 20 psu. In contrast, aCDOM exhibited a more gradual decline and showed less variability than aP. A comparable pattern is evident across the freshwater-to-marine compartments (Fig. 5d). While aCDOM remains relatively stable from the river to the coastal compartment, aP demonstrates a progressive decline across this transition. In the marine compartment, located furthest from the influence of the sediment plume, both particulate and dissolved absorption components show a marked decrease.

Overall, these patterns indicate a systematic shift in the relative contributions of particulate and dissolved optical components along the fluvial-marine continuum. While dissolved organic carbon constituted the dominant fraction of total organic carbon, particulate absorption (aP) contributed disproportionately to total absorption, indicating that particulate components exert a stronger influence on optical properties than suggested by their relative contribution to total organic carbon concentrations.

The distribution of optical water types (OWTs) classified by maximum membership shows a systematic shift along the salinity gradient and across fluvial-marine compartments (Fig. 5e–f). At low to medium salinities (0–20 psu), OWT 6 represents the dominant class, it stands for bright brown water with a high concentration of detritus. In addition, OWT classes 5a and 5b contribute with a green and bright green appearance and a much lower sediment proportion. With increasing salinity, the relative contribution of OWT 5a and 5b increases and becomes more prevalent within the coastal salinity range. At salinities exceeding 20 psu, the sediment component decreases substantially, while classes defined for transitional waters OWT 3a and OWT 2 increase in relative abundance. When grouped by compartment (Fig. 5f), delta samples are primarily associated with OWT 6, whereas coastal waters show a broader distribution of OWT memberships, including a few cases characterized by higher phytoplankton biomass OWT 5a and 5b. The marine compartment is characterized by a shift toward OWT 3a and OWT 2, with only minor contributions from higher-numbered OWT classes. Overall, the distribution of OWT memberships indicates a gradual transition in optical regimes across the fluvial–marine continuum.

A strong, non-linear relationship was found between observed DOC concentrations and aCDOM(443) across the full salinity gradient, with an overall coefficient of determination of r2=0.89 (Fig. 6a). While compartment-specific relationships were generally linear, combining samples across compartments resulted in a non-linear relationship. The overall regression reflects the combined variability of river and delta samples at higher aCDOM(443) values, while marine samples form a tightly clustered low-concentration group. The strength and type of this relationship vary markedly among compartments with the coastal compartment exhibiting the strongest linear relationship (r2=0.86) followed by the delta compartment (r2=0.80). The marine compartment (r2=0.49) exhibited a weaker relationship and narrower range of DOC variability. The compartment specific regressions differed not only in strength but also in slope, with steeper relationships observed in coastal and marine compartments compared to river and delta samples, contributing to the non-linear cross-compartment scaling.

A linear pattern is observed between POC and aP(443) across delta, coastal, and marine compartments (Fig. 6b), with an overall r2=0.71. Note that no aP(443) data was available for the river compartment. Among individual compartments, the marine zone showed the strongest POC-aP(443) relationship (r2=0.91), while the delta compartment exhibited the weakest correlation (r2=0.53). In contrast, the coastal compartment shows a moderate relationship (r2=0.67), indicating spatial variability in the optical-biogeochemical coupling of particulate matter. Slopes of the POC-aP(443) relationships were more consistent across compartments, resulting in a broadly linear cross-compartment relationship despite differences in correlation strength. Samples associated with lower-numbered OWT classes clustered toward lower aCDOM(443) and aP(443) values, whereas higher-numbered OWTs occupied the high-absorption end of the regression space within the subset of samples with available radiometric measurements (Fig. A1). Consistent with the observed POC-aP(443) relationships, POC concentrations exhibited a strong positive linear association with suspended particulate matter (SPM) across the dataset (Fig. A1; r2=0.69, n=162). POC increased linearly with SPM, with variability primarily driven by particle load (Fig. A1).

As riverine freshwater flow transits the interconnected network of distributary channels, it moves through the shallow coastal zone and eventually into the marine zone of the Beaufort Sea. Along this transition, spatial gradients in DOC, POC and SPM observed across compartments (Fig. 3), reflect active chemical and mechanical processing of OC and sediments during transport through the deltaic network and nearshore plume. These processes include flocculation, photochemical and microbial degradation, remineralization, and oxidation, which regulate the fate of terrigenous material during mixing with saline waters (Hernes and Benner, 2003; Vonk et al., 2015; Mann et al., 2016; Bertin et al., 2025; Clark et al., 2022).

While concentration gradients alone cannot uniquely distinguish dilution from biogeochemical processing, shifts in the specific ultraviolet absorbance (SUVA) and the slope ratio (SR) along the salinity gradient indicate changes in DOM composition and molecular character, consistent with progressive photochemical and microbial alteration during transport from river to marine waters (Fig. 7). While the observed shifts in SUVA and SR suggest that the compositional changes in DOM accompany the non-linear DOC-aCDOM(443) relationship, these patterns cannot fully disentangle conservative mixing from in-situ biogeochemical processing, as multiple factors, including overlapping seasonal mixing lines, influence the relationship. Similar spatial gradients in DOC, POC and SPM across riverine, deltaic and shelf environments have been documented for the Mackenzie system and other Arctic deltas, highlighting the role of distributary networks and shallow estuarine environments as active biogeochemical filters that modulate terrestrial carbon export prior to its delivery to the Arctic Ocean (Emmerton et al., 2008; Clark et al., 2022).

At the land-sea interface, where freshwater from the river and delta mixes with saline waters, our observations reveal abrupt shifts in DOC, POC, and SPM distributions, highlighting the mixing zone as a hotspot of material transformation (Figs. 3 and 4). The strong decline in particulate loads and concurrent reorganization of DOC patterns across the plume suggest enhanced aggregation, settling, and selective removal of terrigenous material during early stages of marine exposure. Mixing of riverine freshwater with higher-salinity coastal waters enhances ionic strength and promotes flocculation and aggregation of dissolved and colloidal organic matter, facilitating a transfer from dissolved to particulate organic carbon and enhancing particle settling. Such aggregation-driven phase shifts are well documented in estuarine mixing zones and contribute to the redistribution and removal of terrestrial organic matter within nearshore plume environments (Sholkovitz, 1976; Keil et al., 1997; Burd and Jackson, 2009; Clark et al., 2022).

While flocculation and aggregation are well-documented mechanisms driving DOC to POC conversion and the resulting abrupt concentrations changes in estuarine mixing zones, testing the POC / DOC ratio vs. salinity revealed no clear indication of DOC to POC conversion (Fig. A2). Rather than an increase in the POC / DOC ratio at low salinity conditions (5–10 psu) followed by a continual decrease at higher salinities, a slight increase at moderate salinity (15–20 psu) followed by a gradual decline into saline waters was observed. This pattern suggests that dilution is likely the dominant control with only localized or episodic deviations from flocculation and settling.

Concurrently, as turbidity decreases from the river channel mouth, increased light penetration allows sunlight (UV and visible light) to interact more readily with the surface water. This promotes a breakdown of particulate and dissolved OC via photochemical degradation, oxidation, and remineralization, likely reflected by the 28 % decrease in DOC as waters reach the coastal compartment and an additional 15 % decrease observed in the marine compartment. These estimates are comparable to those reported by Bertin et al. (2025), which suggested that flocculation may account for up to 45 % of DOC loss.

The shallow distributary channels of deltas are typically characterized as areas of sedimentation and deposition. Here, this is reflected by the 38 % decrease in SPM across the river to delta transition (Fig. 3). The much smaller decrease in POC (7 %) and DOC (7 %) across these two compartments highlights the preferential deposition and retention of mineral sediments relative to organic particulates. While some transformation of OC is likely occurring in the delta compartment, these patterns emphasize the dual role of the delta as both a site of retention and transformation. Despite DOC representing the dominant fraction of OC, particulate material exerted a strong influence on optical absorption (Fig. 5c–d), indicating that particle-driven processes shape the optical properties in near-delta environments. This reflects differences in specific absorption efficiency, as suspended mineral and detrital particles exhibit higher absorption and scattering cross-sections than dissolved organic matter (Doxaran et al., 2012). Processes such as primary production and in-situ POC generation via phytoplankton growth may also occur in deltaic waters, although are likely limited in these highly turbid waters due to reduced light penetration (Clark et al., 2022).

Similar to the Yukon River Delta and Lena River Delta systems (Juhls et al., 2019; Clark et al., 2022), DOC concentrations in the Mackenzie River Delta exhibited a distinct spring freshet peak (Fig. 2), followed by moderately elevated levels that persisted through the ice-free season. This pattern suggests that snow and ice-melt driven inputs during breakup provide an initial DOC pulse, while continued mobilization from catchment soils and small tributaries sustains concentrations into summer. In contrast, POC and SPM showed more episodic variability, with significantly high freshet peaks followed by secondary peaks during mid-summer high-discharge events (Figs. 3c and e; A1a and c). These dynamics highlight the combined influence of seasonal hydrology and local deltaic processes, or catchment disturbance processes, such as permafrost thaw bank erosion and thermokarst lake dynamics (Douglas et al., 2023).

Previous studies from Arctic river-shelf systems have shown that temporally constrained river-plume transects can exhibit near-conservative mixing relationships between DOC, CDOM, and salinity (e.g., Grunert et al., 2021; Hölemann et al., 2021; Clark et al., 2022). However, due to the complex dynamics and biogeochemistry of the seasonal freshet, more intensive sampling during and between under-studied high flow conditions is needed to improve our understanding of OC composition and reactivity across the land-ocean scale (Burns et al., 2024).

When combining observations from multiple seasons, years, and spatial compartments, overlapping seasonal mixing trajectories may produce more diffuse or non-linear annual relationships. Figure 4a demonstrates the influence of the seasonal spring freshet on DOC conservative mixing, where observations collected during the May–June freshet result in a freshwater endmember of 6.61 mg L⁻¹ compared to the year-round mean freshwater endmember (5.21 mg L⁻¹). Similarly, aCDOM(443) illustrates a similar relationship along the salinity gradient (Fig. 4b), where measurements collected during the spring freshet shift the theoretical mixing lines, which is visible by the changing slopes (-0.06 m⁻¹ psu⁻¹ during the May–June freshet to -0.04 m⁻¹ psu⁻¹ including year-round observations).

Since processes such as degradation, flocculation, coastal erosion, sea-ice meltwater input, and seasonal variability in freshwater endmembers overlap spatially and temporally, it remains difficult to attribute deviations from conservative mixing to a single mechanism. We therefore interpret the observed patterns as reflecting a combination of conservative mixing and seasonally variable biogeochemical processing across the fluvial–marine continuum.

Salinity-DOC and salinity-aCDOM(443) relationships reveal a narrow zone of non-conservative behavior at low salinities (0–5 psu), where DOC concentrations decline more steeply than predicted by conservative mixing. This pattern is consistent with partial transformation or removal of chromophoric DOM during the initial freshwater-marine transition. At salinities above 5 psu, both DOC and aCDOM(443) decrease linearly with salinity indicating predominantly conservative transport through the inner plume, in agreement with Juhls et al. (2022) (Fig. 4a and c). The observed decline in aCDOM(443) from 3.2 to 1.8 m⁻¹ within 0–5 psu likely reflects a combination of processes and compositional shifts in DOM.

Overall, these results indicate that most terrigenous DOC is transported conservatively through the inner plume, while a measurable fraction is altered or removed near the freshwater-marine interface. In contrast, POC and SPM both exhibited a strong decrease at low salinities, indicating distinct non-conservative behavior associated with particle settling (Fig. 4). The variation observed in particulate concentrations near 15 psu could be reflective of resuspension and redistribution processes that can occur during offshore transport.

These transformations can have implications for air-sea CO₂ exchange. Terrestrial OC fluxes strongly influence CO₂ exchange in coastal waters (Nielsen et al., 2022). Recent estimates suggest that OC inputs reduce the AO's net CO₂ sink strength by 7 %–14 % (Nielsen et al., 2022) and may convert some shelves into net CO₂ sources (Bertin et al., 2023). Our observations indicate non-conservative behavior and partial removal of DOC within the near-delta plume, deviating from a linear mixing line between freshwater and saline endmembers, suggesting that roughly 27 % of riverine DOC may be transformed or lost at low salinity coastal waters. Scaling this first-order estimate to the mean Mackenzie River DOC export (∼ 1.4 Tg C yr⁻¹; Matsuoka et al., 2022) implies a potential removal of ∼ 0.38 Tg C yr⁻¹ within the mixing zone. Assuming a fraction of 10 %–15 % of this removed DOC undergoes photochemical mineralization (Bertin et al., 2023), this corresponds to an estimated CO₂ release of 0.04–0.06 Tg C yr⁻¹. While approximate, these values underscore that DOC transformation in the low-salinity plume can represent a measurable contribution to coastal carbon outgassing.

The compartmental framework applied here provides a system-scale context for interpreting OC and sediment transport and retention across the Arctic land-ocean continuum (Fig. 8). The river, delta, coastal, and marine compartments reveal distinct regimes of organic carbon processing, reflected in the observed gradients in concentration, optical properties and bio-optical relationships (Figs. 3–6). By resolving these four distinct compartments across the entire salinity gradient, our results highlight how hydrodynamic sorting, mixing behaviors and optical regime shifts collectively shape OC dynamics in Arctic river-shelf environments. While similar patterns have been reported in other observations from Arctic river-shelf systems (Juhls et al., 2019, 2022; Clark et al., 2022), here they are resolved consistently across all major components of the fluvial–marine continuum.

To address uncertainties in satellite retrievals for optically complex Arctic coastal waters, we derived compartment-specific and cross-compartment bio-optical relationships spanning the full salinity gradient (Fig. 6). Within individual compartments, DOC-aCDOM(443) relationships were largely linear, whereas combining data across compartments resulted in a distinctly non-linear relationship. Similarly, POC-aP(443) relationships varied among compartments, reflecting differences in particle composition and optical regime along the continuum. These contrasts underscore that riverine, deltaic, coastal, and marine waters exhibit distinct optical and biogeochemical coupling, emphasizing the need for retrieval approaches that account 555 for spatial variability rather than relying on a single pan-arctic parameterization.

When examined across the full dataset, the non-linear DOC-aCDOM(443) relationship (Fig. 6a) reflects systematic changes in the composition and optical quality of DOM along the fluvial–marine continuum. In the deltaic and low-salinity coastal zones, mixing of terrestrially derived, highly absorbing DOM with less colored marine DOM reduces CDOM absorption relative to DOC concentrations, flattening the relationship. Photobleaching and microbial processing during offshore transport preferentially remove chromophoric components relative to the total DOC (Vonk et al., 2015). As a result, relationships that appear linear within individual compartments as reported in many studies (e.g., Mann et al., 2012; Walker et al., 2013; Novak et al., 2022) diverge non-linearly when considered across the full cross-compartment continuum, which integrates these compositional and optical shifts (Fig. 6a; black solid line).

Previous studies from Arctic river-shelf systems have shown that temporally constrained river–plume transects can exhibit near-conservative mixing relationships between DOC, CDOM, and salinity (e.g., Clark et al., 2022; Hölemann et al., 2021; Grunert et al., 2021). However, due to the complex dynamics and biogeochemistry of the seasonal freshet, more intensive sampling during and between under-studied high flow conditions is needed to improve our understanding of OC composition and reactivity across the land-ocean scale (Burns et al., 2024).

When combining observations from multiple seasons, years, and spatial compartments, overlapping seasonal mixing trajectories may produce more diffuse or non-linear annual relationships. Since processes such as degradation, flocculation, coastal erosion, sea-ice meltwater input, and seasonal variability in freshwater endmembers overlap spatially and temporally, it remains difficult to attribute deviations from conservative mixing to a single mechanism. We therefore interpret the observed patterns as reflecting a combination of conservative mixing and seasonally variable biogeochemical processing across the fluvial–marine continuum.

While compartment-specific bio-optical relationships have commonly been used for regional retrievals (Doxaran et al., 2015; Matsuoka et al., 2013, 2017), our analysis shows that such relationships may not be transferable across the full salinity gradient. The emergence of non-linear cross-compartment behavior and changing slopes of the relationships with adding compartments (Fig. 6) suggests that synoptic-scale remote sensing requires relationships derived across the entire continuum of optical conditions.

Seasonal contrasts between the ice-free period and the full annual dataset further refine these patterns. Cross-compartment relationships between DOC-aCDOM(443) strengthened during the ice-free season (filtering data to only July–October), with the coefficient of determination increasing to r2=0.92, whereas the POC-aP(443) relationship remained comparatively stable from r2=0.71 annually to r2=0.72 in the ice-free season (Fig. 6). Stronger coupling during the ice-free season likely reflects more consistent optical regimes, whereas winter freshwater samples add variability. Winter in-situ observations were largely limited to freshwater compartments, as sea-ice cover restricted access to coastal and marine compartments. The spring freshet (May–June) represents an important but under-sampled period in this seasonal cycle, when large pulses of freshwater transport elevated OC and sediment concentrations into the system.

Although overall seasonal differences in bio-optical relationships were modest, continued monitoring remains essential, as Arctic warming and hydrological intensification are expected to increase variability in the amount and composition of CDOM and suspended particulates (Bertin et al., 2025; Novak et al., 2022; Chalov et al., 2023).

Our observations confirm that optical properties in Arctic coastal waters are strongly controlled by both CDOM and suspended particles, with their relative importance varying among compartments (Fig. 5c–d). The enhanced absorption in blue wavelengths (400–500 nm) produces the characteristic brown–green coloration of deltaic and nearshore waters. In low-salinity deltaic waters, aP exceeded aCDOM by a factor of ∼ 1.5–2, indicating that mineral and detrital particles dominate light attenuation despite DOC representing the major fraction of OC. This reflects differences in mass-specific absorption and scattering efficiency. These shifts illustrate how optical–biogeochemical coupling evolves across the fluvial–marine continuum where particle-rich waters obscure CDOM signals and complicate retrievals of DOC from remote sensing. Photodegradation and compositional changes in CDOM further modulate these gradients (Ward et al., 2017; Grunert et al., 2021; Clark et al., 2022; Moran et al., 2022), underscoring the need for cross-compartment, or OWT-specific bio-optical algorithms (Fig. A3) in Arctic coastal and deltaic environments.

While bio-optical relationships refine the retrieval of DOC and POC, OWT classification provides an additional framework for interpreting optical variability and guiding algorithm selection in Arctic coastal waters. In our dataset, OWT distributions revealed a structured transition from OWT 6 dominance in river and delta compartments toward increased contributions of OWT 4a–4b in coastal waters and OWT 3a–2 in marine environments (Fig. 5f). Previous studies have associated higher-numbered OWT classes with stronger particulate and detrital scattering signatures, whereas lower-numbered classes typically reflect increasing contributions of phytoplankton and clearer water optical conditions (Bi and Hieronymi, 2024). The observed progression therefore suggests a shift from particle-rich delta waters toward optically clearer offshore regimes, consistent with the measured changes in aP and aCDOM along the fluvial–marine continuum.

Within the delta compartment, localized occurrences of OWT 4a and 4b likely reflect reduced hydrodynamic connectivity and enhanced particle settling in smaller distributary channels and floodplain environments. Such conditions may promote changes in the relative contributions of detrital material, phytoplankton biomass, and CDOM, resulting in optical regimes that differ from those of the main plume pathway. By linking reflectance-based OWT classes to dominant optical regimes, OWT classification provides a practical basis for selecting regime-specific retrieval approaches in ocean color remote sensing (OCRS) (Fig. A3). Incorporating OWT-dependent parameterizations may therefore reduce retrieval bias and improve estimates of carbon and sediment dynamics across optically complex river-sea systems (Hieronymi et al., 2026).

Together, the observed shifts in optical regimes (Fig. 5) and the emergence of non-linear cross-compartment bio-optical relationships (Fig. 6) demonstrate how land-sea connectivity controls both the biogeochemical and optical structure of Arctic coastal waters. With continued warming, enhanced runoff, permafrost thaw, and coastal erosion are likely to further intensify sediment and OC delivery, potentially amplifying variability in optical conditions across the fluvial-marine transition zone. Integrating spatially resolved in-situ observations with OWT-informed remote sensing approaches offers a pathway to capture these transitions at synoptic scales, improving the robustness of DOC, POC, and SPM retrievals in optically complex Arctic environments. These findings highlight that accurate monitoring of terrestrial matter transport on Arctic shelves requires frameworks that account for evolving optical regimes rather than relying on static regional calibrations.