A systematic review of the methods and magnitudes of dissolved oxygen exchanges in streams and rivers

1 Introduction

Dissolved oxygen (DO) is a master variable in aquatic ecosystems as it integrates physical, biological, and chemical processes occurring within a waterbody and the connected river corridor (Hutchinson, 1938). DO serves as a constraint on aquatic life, an indicator of ecosystem processes and their dynamics (e.g., Davis, 1975), and regulates the cycling of both macro- and micronutrients (e.g., Davison, 1993). In well-oxygenated environments, microbial respiration uses oxygen as a terminal electron acceptor, but as DO is depleted in space and time, processes such as nitrate, iron, and sulfate reduction emerge (e.g., Morrice et al., 2000). These redox transitions influence the production of greenhouse gases (e.g., McClain et al., 2003; Beaulieu et al., 2011), mobilization of contaminants (e.g., Borch et al., 2010; O’Donnell et al., 2024), and determine the net release or retention of nutrients (e.g., Hupfer and Lewandowski, 2008). At the same time, DO levels can constrain biological communities, where low oxygen (<5 mg L⁻¹) can impair growth and reproduction in fish and macroinvertebrates (e.g., Vaquer-Sunyer and Duarte, 2008; Saari et al., 2018). Moreover, hypoxia (<2 mg L⁻¹) has been increasingly documented in aquatic ecosystems globally, driven by a combination of climate warming, nutrient enrichment, and hydrologic alteration (Jane et al., 2021; Blaszczak et al., 2023; Chan et al., 2025). While we have extensive knowledge about DO concentrations in lotic systems, we posit that our present understanding remains limited by methodological heterogeneity, divergent scales of investigation, and disciplinary differences in measurements priorities, modes of inference, and nomenclature.

In most lotic systems, physical (Zhi et al., 2023) and biological (Garvey et al., 2007) controls overwhelmingly dominate oxygen budgets, while chemical processes play a comparatively minor role. Odum (1956) formalized this in an oxygen mass balance equation:

d O 2 dt = GPP − ER + k ( C s − C )

where d O 2 dt is the net areal oxygen flux (g O₂ m⁻² day⁻¹), k is the gas transfer velocity (m day⁻¹), C s is the dissolved oxygen concentration at saturation (g O₂ m⁻³), and C is the in-stream dissolved oxygen concentration (g O₂ m⁻³). Building on this early oxygen balance, we organize oxygen exchanges—the processes through which oxygen enters or leaves stream compartments—into five distinct process domains. First, whole-stream metabolism (GPP and ER) is an integrative measure of stream DO that emerges from the interaction of multiple underlying processes. Next, these underlying processes are grouped by where oxygen is added or removed within (1) the water column, (2) the benthos, (3) the hyporheic zone and sediments, and (4) surface-atmosphere exchange (Figure 1). Together, these four groups represent distinct yet coupled physical and biological processes that regulate the balance of oxygen, the net aggregation of which manifest as whole-stream DO concentrations and diel dynamics. A comprehensive understanding of DO and its dynamics therefore requires bridging disciplines and linking exchanges across spatial and temporal scales, as compartment dominance may vary between stream settings and shift seasonally or under disturbance (Genzoli and Hall, 2016).

Our objective is to critically evaluate how oxygen exchanges and their controls have been studied in streams and rivers. Specifically, we ask: to what extent does the existing literature capture the interconnected exchanges that comprise a stream oxygen budget, and how have these exchanges been studied? To address this, we document the methods used to quantify oxygen exchanges and how they have evolved over time, which exchanges have been measured along with their magnitude and variability relative to whole-stream fluxes, how often exchanges have been quantified concurrently, and the metadata reported alongside exchange measurements. Our results highlight biases and limitations in the literature, from which we outline key challenges to achieving a holistic understanding of DO dynamics in streams and propose a path toward building integrated stream and river oxygen budgets.

2 Defining oxygen exchanges in streams and rivers

2.1 Whole-stream metabolism

Gross primary production (GPP) is inferred from the total oxygen production via photosynthesis, while ecosystem respiration (ER) reflects the total oxygen consumption by autotrophs and heterotrophs; their difference (i.e., GPP-ER) defines net ecosystem production (NEP), capturing the balance between autotrophic energy fixation and aerobic respiratory demand (Odum, 1956). In forested headwaters, shading, low nutrients, and allochthonous organic matter subsidies allow ER to exceed GPP, whereas downstream reaches with greater light availability or elevated nutrients typically have higher GPP, generating predicted longitudinal patterns of NEP in stream and river networks (e.g., Vannote et al., 1980).

Spatial patterns in river network metabolism are well-established (e.g., Vannote et al., 1980; Mulholland et al., 2001; Mejia et al., 2019). Most whole-stream metabolism models and empirical estimates are done at the reach-scale, where accuracy improves with longer time series and well-constrained gas exchange (Appling et al., 2018a). However, scaling reach-scale metabolism estimates to the network or basin remains challenging due to equifinality among parameters, site-specific characteristics, and event-driven disturbance that disrupt diel patterns including sediment pulses, storms, and flow intermittence (Bernot et al., 2010; Appling et al., 2018a; O’Donnell and Hotchkiss, 2022; López-Rojo et al., 2025). In response, a growing body of literature is advancing network-scale modeling frameworks (e.g., Koenig et al., 2019; Segatto et al., 2023) and synthesizing metabolic patterns that emerge across regions (Bernhardt et al., 2022). Emerging work further integrates stream metabolism with process-based approaches such as hydraulic routing (Payn et al., 2017), partitioning autotrophic and heterotrophic production (Hall and Beaulieu, 2013), and linking metabolic rates to CO₂ cycling and emissions (Hotchkiss et al., 2015). Despite these advances, challenges remain when interpreting DO time series measured in surface waters, as most metabolism models assume that observed DO dynamics reflect in-stream biological processes and surface-atmosphere exchange alone. In reality, streams lose and gain oxygen from multiple hydrologic pathways including hyporheic exchange, groundwater inputs or losses, dynamic channel storage, and overland flow, each carrying DO concentrations that may differ from stream water and produce signals that can be misattributed to biological production or consumption (e.g., Hall and Tank, 2005; Gomez-Velez et al., 2026).

Early diel DO measurements to assess whole-stream GPP and ER were conducted manually using Winkler titrations, a labor-intensive approach that limited temporal resolution and duration. High-frequency in situ sensors have revolutionized the estimation of GPP and ER by allowing continuous monitoring and reach-scale integration (Marzolf et al., 1994). The most widely used approach for quantifying GPP and ER is the diel oxygen technique, or whole-stream metabolism, where single- and two-station models solve oxygen mass balances to estimate GPP, ER, and gas exchange simultaneously (Riley and Dodds, 2013; Grace et al., 2015; Hall et al., 2016). Recent advances in modeling tools, such as streamMetabolizer, have improved parameter estimation using Bayesian inference, by allowing gas exchange to vary with discharge, examining DO residuals, reducing equifinality, and enabling meta-analysis of field studies (Holtgrieve et al., 2010; Appling et al., 2018a; Appling et al., 2018b).

While whole-stream estimates of GPP and ER have advanced our understanding of DO and carbon dynamics at the scale of river reaches and networks, improved understanding requires quantifying the process domains that contribute to integrated whole-stream observations (sections 2.2–2.5). Resolving these exchanges is essential because oxygen dynamics arise from the superposition of the water column, benthos, hyporheic zone and underlying sediments, and surface-atmosphere exchange, the combination of which can create signals that obscure inferences of whole-stream metabolism (Figure 1).

2.2 Water column production and respiration

Water column metabolism refers to photosynthesis and respiration occurring within the water column, primarily by free-floating phytoplankton and microbes and communities on suspended particles and organic matter. Water column contributions to whole-stream metabolism increase broadly with channel size, where greater depth, residence time, and light availability favor planktonic communities (e.g., Vannote et al., 1980; Yang et al., 2021). While minimal in most low-order streams, water column metabolism can represent a substantial fraction of whole-stream metabolism in mid-sized, nutrient-enriched rivers (Oliver and Merrick, 2006; Reisinger et al., 2016). Early methods relied on light–dark bottle incubations and in situ chambers, which isolated water column processes without the confounding effect of gas exchange (Hargrave, 1972; Bott et al., 1978). While these techniques remain useful, they are labor-intensive and capture only a snapshot in time. More recent empirical studies have demonstrated how disturbance can shift the balance between water column and benthic metabolism at the reach-scale (Genzoli and Hall, 2016), and integrative approaches have been used to link water column metabolism to broader biogeochemical cycling (e.g., nutrient uptake and denitrification) across land use gradients (Reisinger et al., 2021). Although emerging technologies offer new pathways, they face several challenges, including the high cost and limited suitability of autonomous optode-based sensors for smaller streams (Collins et al., 2018), optical complexity and riparian canopy cover that constrain remote sensing of phytoplankton (Spyrakos et al., 2018), and difficulties scaling process-based approaches developed for lakes across river networks (Van De Bogert et al., 2007). Concurrent water column measurements are therefore essential for interpreting whole-stream metabolism signals, particularly in mid-sized and larger rivers where planktonic processes can represent a substantial and seasonally variable fraction of whole-stream metabolism.

2.3 Benthic production and respiration

Benthic metabolism encompasses photosynthesis and respiration occurring on the streambed, driven by biofilms, periphyton, macrophytes, and sediment microbial communities. In most streams and rivers, benthic processes dominate relative to the water column, though their contribution varies with land use and disturbance (Rutherford, 1991; Hall et al., 2003). Benthic metabolism is typically measured using chambers that enclose patches of the streambed (Mulholland et al., 2001; Rüegg et al., 2015; Lopez et al., 2025). Partitioning contributions of different benthic materials has been a research focus with studies targeting macrophyte production (Fisher and Carpenter, 1976; Hensley and Cohen, 2020), algal photosynthesis (Mulholland et al., 1994; Acuña et al., 2011), and microbial respiration in sediments (Hill et al., 2000). Recently, aquatic eddy covariance—a technique widely applied in wetlands and marine systems—has been applied to measure in situ benthic oxygen fluxes in streams (Rovelli et al., 2017), though these measurements integrate turbulent exchange near the bed and cannot fully separate benthic, hyporheic, and water column contributions. Scaling benthic metabolism in space is difficult due to sediment heterogeneity (Lee-Cullin et al., 2018), with fluxes varying up to two orders of magnitude depending on substrate type, catchment land use, disturbance regime (Fellows et al., 2006), and organic matter quantity (Hill et al., 2017). This variation implies that efforts to quantify these fluxes, especially when estimates are conjoined with whole-stream measurements, will improve our understanding of how whole-system fluxes are partitioned.

2.4 Hyporheic respiration and sediment oxygen demand (SOD)

The hyporheic zone is the interface within stream sediments where surface water exchanges with shallow groundwater, encompassing flowpaths where water temporarily leaves the stream channel and undergoes biogeochemical transformation before returning (White, 1993). Hyporheic respiration is the oxygen consumption along these subsurface flowpaths and is driven primarily by microbial and faunal respiration, with its impact on stream DO controlled by residence time, exchange flux, and reaction rates (Gomez et al., 2012). Sediment oxygen demand (SOD) is conceptually related—both quantify oxygen consumption in the stream-connected subsurface—but the two diverge between study purpose and method. SOD emerged as a stream-centric concept to explain oxygen deficits not accounted for by in-stream processes alone, particularly in the context of predicting oxygen sag and organic loading along rivers (Thomann, 1972; Lee et al., 1978). SOD is typically quantified using static cores that do not capture the full complexity of in situ transport processes (Bowman and Delfino, 1980; Coenen et al., 2019). In contrast, hyporheic research is largely focused on understanding subsurface ecosystem processes (Ward, 2016).

Hyporheic respiration has been quantified by vertical oxygen profiles, pore water sampling, and in situ or ex situ incubations (Revsbech et al., 1986; Uzarski et al., 2001; Nelson et al., 2019). Experimental flumes have been used under controlled flow and sediment conditions to isolate hyporheic exchange and determine how reaction rates and biogeochemical processes vary across redox conditions (Reeder et al., 2018), while streamside mesocosms packed with bed sediment have been used to replicate in situ flow paths to the maximum extent practicable (Serchan et al., 2024; Herzog et al., 2025). High-resolution DO sensors have been deployed directly into hyporheic sediment to measure oxygen along hyporheic flowpaths through time (Malcolm et al., 2006; Gooseff et al., 2023), complemented by dissolved gas time series that reveal hyporheic exchange and redox transitions (Mächler et al., 2013). These advances coincided with increasing use of dimensionless transport-reaction frameworks such as Damköhler numbers and the Reaction Significance Factor to evaluate the balance between hydrologic transport and biogeochemical reaction rates within hyporheic flowpaths (e.g., Gomez-Velez and Harvey, 2014; Harvey et al., 2019).

Because these subsurface fluxes are central to both the oxygen budget of a stream, and to the genesis of open-channel DO signals (Gomez-Velez et al., 2026), renewed attention to quantifying heterogeneous subsurface flowpaths and processes as part of whole-stream measurements is likely to be critical to improved understanding of metabolic variation in time and space.

2.5 Surface-atmosphere exchange

Surface-atmosphere exchange, or gas exchange, is the transfer of dissolved gases across the air-water interface, governed by turbulence as well as water temperature, gas solubility, and the concentration gradient between the water and atmosphere (Wanninkhof, 1992; Melching and Flores, 1999; Raymond and Cole, 2001; Ulseth et al., 2019; Hall and Ulseth, 2020). The study of gas exchange in rivers has advanced along three primary methodological pathways. First, empirical tracer injections using a specific gas tracer (e.g., propane, SF₆, Ar) are used to measure degassing and inform empirical (e.g., Schmidt-number) relationships (Tsivoglou and Wallace, 1972; Rathbun et al., 1978; Genereux and Hemond, 1992). However, scaling gas transfer coefficients across different gases can introduce error under bubble-mediated conditions because of differences in gas solubility and diffusivity (Klaus et al., 2022). To minimize scaling uncertainty, argon has emerged as an ideal tracer due to its similarity in solubility and diffusivity to oxygen (Hall and Madinger, 2018). Floating dome chambers offer a direct, non-invasive, and easily scalable method for measuring gas exchange, avoiding the logistical challenges of gas tracers, but also provide smaller-scale measurements prone to bias when deployments alter stream turbulence (Lorke et al., 2015). Second, analyses of DO time series are used to infer gas exchange. The most widely used techniques include (1) nighttime regression, which uses the nighttime decline in DO after photosynthesis ceases to estimate respiration, requiring sufficient daytime oxygen production to generate a resolvable nighttime slope (Hornberger and Kelly, 1975); (2) the solar noon lag, which examines the temporal offset between peak light and peak DO (McBride, 2002); and (3) the delta method, which quantifies fluxes by partitioning observed changes in DO storage during both day and night (Chapra and Di Toro, 1991). Each of these approaches relies on detectable diel signals generated by the dynamics of photosynthesis and performs best in medium to large rivers where turbulence is lower and oxygen dynamics are less spatially heterogenous (Bailey and Ahmadi, 2014). Finally, based on tracer injection and other field studies, empirical relationships to predict gas exchange as a function of river hydraulics allow for gas exchange estimates without direct measurement, although they may not capture site-specific conditions (Hall and Hotchkiss, 2017). Early models related gas exchange rates to velocity and depth (O’Connor and Dobbins, 1958), while later work incorporated turbulence metrics such as the energy dissipation method (Tsivoglou and Neal, 1976; Chu and Jirka, 2003), and used prior tracer experiments to develop physical models of gas exchange for small streams (Raymond et al., 2012; Aho et al., 2025). More recently, these models have been applied in scaling gas exchange across river networks and are widely applied in regional and global syntheses (Raymond et al., 2013; Hotchkiss et al., 2015).

The predictability of gas exchange is scale- and system-dependent regardless of which approach is used. Despite their utility, hydraulic equations are fundamentally reach-averaged and are typically derived from log-transformed data, resulting in high predictive uncertainty in morphologically complex headwaters, where turbulence and bubble entrainment vary over short spatial scales (Rovelli et al., 2018; Vautier et al., 2020). Consequently, site-specific calibration or empirical measurements remain essential for accurate flux estimates in many streams and rivers (Hall and Ulseth, 2020). Methodological advances including acoustic Doppler velocimetry (Gålfalk et al., 2013), stable isotopes (Tobias et al., 2007; Jamieson et al., 2013), and sensor-equipped floating chambers (Vachon et al., 2010; Arce et al., 2021) have increased our ability to resolve gas fluxes at higher spatial and temporal resolution, yet scaling these approaches across heterogeneous systems remains a significant challenge (Botter et al., 2022). We note that the goal of empirically constraining gas exchange to ensure robust metabolism estimates has been widely recognized, with conjoined measurements of k₆₀₀ and open-channel DO time series now serving as method best practice (Appling et al., 2018a). This recognition highlights our central premise that attention to other process domains will likely enhance the validity of metabolism measurements.

3 Methods

Despite decades of research, DO in streams and rivers is most often characterized as the integrated, whole-stream signal, which provides valuable information on oxygen and carbon fluxes. However, where those signals are being generated (water column, benthos, hyporheic zone and sediments, and surface-atmosphere exchange) is less often characterized. Here, we conduct a systematic review to document the methods used to measure oxygen exchanges across stream compartments, their magnitudes and variability, and the extent to which process domain fluxes have been measured concurrently with whole-stream metabolism, with the goal of identifying advances and persistent gaps on the path toward an integrated stream oxygen budget.

To assess the extent to which oxygen exchanges and methods have been applied in the literature, we conducted a search on Web of Science on August 26, 2024, using the keywords [“stream” OR “river” OR “aquatic”] AND [“dissolved oxygen” OR “O₂”], yielding 4,086 studies (Figure 2; Supplementary Table S1). The top fields of study represented in the body of literature were environmental sciences (1,487), water resources (622), and environmental engineering (578). After excluding synthesis papers and non-stream or river systems (e.g., lakes, wetlands, marshes, and saltwater systems), 2,832 articles were screened by title and abstract. At this stage, we applied a broad inclusion criterion, retaining any study that reported DO or oxygen-related water quality parameters in a stream or river, regardless of whether a quantified exchange was explicitly reported. Full-text review of the retained 914 studies applied stricter criteria, requiring that studies (1) included empirical data and (2) reported a quantified oxygen exchange as a rate or flux. We define oxygen exchanges broadly as all quantified transfers of oxygen into or out of stream compartments, distinguishing a flux as mass per unit area per unit time and a rate as mass per unit time (after Stauffer, 2006; Supplementary Table S2). Full-text review yielded 204 papers meeting both criteria. Because only 8 of these included explicit hyporheic observations, we relaxed inclusion criteria for hyporheic-centric studies to retain those reporting subsurface DO concentrations or empirical mesocosm work as a proxy for hyporheic conditions, yielding an additional 26 papers and a final synthesis of 230 studies (Figure 3a). Studies were geographically skewed toward the U.S. (51%), followed by Germany (7%), the United Kingdom (6%), and limited representation from Africa, Northern Eurasia, and the southern hemisphere (Figures 3b,c).

For each study, we recorded geographic location, study period, stream size, land use, watershed attributes, hydrologic metrics, flow regime (perennial or intermittent), surface water chemistry, subsurface characteristics, methods used, and oxygen rate or flux measurements (see dataset for more details). For comparison of exchange magnitudes, we analyzed only metabolic exchanges standardized to areal fluxes (g O₂ m⁻² d⁻¹), omitting 25 studies without reported areal fluxes, and analyzed gas exchange as k₆₀₀ (d⁻¹) normalized to a temperature-dependent Schmidt number (including n = 64 studies for which we computed k₆₀₀ from reported data; Supplementary Figure S1).

To evaluate how oxygen exchanges have been measured, we classified methods into eight categories (Box 1): (1) tracers—conservative gas additions (e.g., Ar, SF₆, propane) and isotope tracers (e.g., ¹⁶O, ¹⁸O) that quantify oxygen exchange and transport; (2) in-stream DO—discrete in situ measurements (e.g., porewater profiles) and continuous sensor deployments that estimate GPP, ER, and gas exchange from diel DO dynamics; (3) empirical equations—velocity-depth or shear stress equation predicting gas exchange coefficients from hydraulic variables; (4) interface flux—methods that directly measure oxygen flux or turbulence at the air-water or sediment–water interface without enclosing the system, including aquatic eddy covariance, floating probes with dissolved oxygen sensors, reactive tracers (e.g., oxalic acid), and sound velocity approaches (Koopmans and Berg, 2015; De Souza Ferreira et al., 2020); (5) standing stock—accumulated biomass or chlorophyll-a as a proxy for metabolic activity; (6) field chambers—in situ benthic incubations, mesocosms, and light–dark bottles; (7) lab chambers—sediment or mesocosm incubations conducted under controlled laboratory conditions; and (8) process-based models—coupled biogeochemical (e.g., OTIS, PFLOTRAN) and water quality models (e.g., QUAL2K, WASP) retained only when paired with empirical observations.

Definitions of oxygen exchanges and their typical methods and scales.

Oxygen dynamics in streams emerge from processes operating across distinct compartments and spatial scales, each quantified using different measurement approaches. As summarized below, these fluxes integrate processes over spatial domains, spanning centimeter-scale sediment measurements to reach-scale estimates of metabolism and gas exchange, highlighting the ongoing challenge of integrating oxygen exchanges across compartments and methodology.

Sediment oxygen demand: Total oxygen consumption by streambed sediments due to biological respiration and abiotic oxidation of reduced compounds

• Typical methods: Quantified using ex situ sediment cores or in situ chambers
• Typical spatial scale: Patch scale (centimeter to meter)

Benthic metabolism: Photosynthesis and respiration occurring on the streambed, often by biofilms, periphyton, macrophytes, and sediment microbial communities

• Typical methods: Measured using in situ benthic chambers or incubations of rocks or sediments
• Typical spatial scale: Patch scale (centimeter to meter)

Water column metabolism: Photosynthesis and respiration occurring within the water column, often by phytoplankton and microbes

• Typical methods: Estimated using in situ chambers such as light–dark bottles
• Typical spatial scale: Patch scale (decimeter to meter)

Hyporheic respiration: Biologically mediated oxygen consumption along subsurface flow paths where surface water exchanges with shallow groundwater

• Typical methods: Quantified using in situ DO, in situ hyporheic chambers, or inferred from residence-time modeling, and tracer- or time-series approaches (e.g., argon, conductivity)
• Typical spatial scale: Patch scale (centimeter to meter) to sub-reach (1–10 m)

Gross Primary Production (GPP) and Ecosystem Respiration (ER): Total oxygen produced via photosynthesis and consumed by autotrophs, heterotrophs, sediments, and subsurface pathways

• Typical methods: Inferred from diel DO signals using diel metabolism models or in situ chambers
• Typical spatial scale: Reach-scale (10 s of meters to kilometer)

Surface-atmosphere exchange: Physical exchange of oxygen between the stream and the atmosphere

• Typical methods: Estimated using in situ DO time series, tracer gas injections (e.g., SF₆, propane, argon), empirical equations, or floating probes
• Typical spatial scale: Reach-scale (10 s of meters to kilometer)

4 Results

4.1 How are oxygen exchanges being measured?

Across oxygen exchanges, methodological differences reflected both the spatial scale of measurement and the process of interest (Box 1). In-stream DO was the most common approach for quantifying GPP (81% of studies), ER (79%), gas exchange (49%), and hyporheic respiration (46%), reflecting both discrete and continuous measurements that have evolved through time (Figure 4). DO-based estimates of hyporheic respiration often relied on the difference between whole-stream and benthic metabolism rather than being directly measured (e.g., Naegeli and Uehlinger, 1997; Acuña et al., 2011; see section 5.3 for limitations). After in-stream DO, gas exchange was most commonly quantified using tracers (25% of studies), encompassing a variety of gases—including argon, propane, helium, xenon, ethylene, nitrogen, sulfur hexafluoride, and krypton (e.g., Young and Huryn, 1999; Roley et al., 2023)—with additional tracers used to link gas exchange and metabolic rates (Piatka et al., 2021; Barbosa et al., 2023), followed by empirical equations (20%). In contrast, water column, benthic, and SOD measurements were dominated by chamber-based approaches, with field chambers accounting for 83, 64, and 12% of studies and lab chambers comprising 11, 24, and 88%, respectively.

Beyond in-stream DO, tracers, and chamber-based approaches, a suite of less frequently applied methods has been used to quantify oxygen exchanges. Interface flux approaches have expanded exchange measurements across flow regimes by capturing near-continuous gradients in situ such as the soluble floating probe method (De Souza Ferreira et al., 2020), floating probes with DO sensors (Hogsett and Goel, 2013), and aquatic eddy covariance (Koopmans and Berg, 2015). The interface flux category was highest for benthic exchanges (8% of studies), then gas exchange (6%), and GPP/ER (1%). Other approaches, including standing stock estimates, were applied exclusively to GPP and ER (1%). Lastly, process-based modeling studies informed by empirical observations, including frameworks such as the PROSE model (Even et al., 2004) or reactive transport models (Diem et al., 2013) accounted for 18% of hyporheic, 6% of water column, 4% of benthic, and less than 1% of GPP, ER, and gas exchange studies.

Methods to quantify exchanges have also shifted through time, both in frequency and diversity (Figure 4). After relatively few studies in the 1960s, research on most oxygen exchanges increased in the 1970s, except for hyporheic exchange and SOD, which peaked in the 2010s. The apparent decline in the number of studies after 2020 reflects limitations of our analysis only considering studies with DOIs assigned before August 2024. Field chambers dominated early studies, comprising all studies in the 1960s and about half in the 1970s, but declining to 8% by the 2010s. In contrast, in-stream DO methods have greatly increased in number and gradually increased by percentage from 46% in the 1970s to 56% in the 2020s, reflecting the transition from discrete sampling to continuous sensor deployments as sensor technology advanced (Rode et al., 2016). Whole-stream metabolism methods have remained relatively stable over time, with single-station approaches dominating, while methodological diversity increased in the 1990s–2000s with greater use of two-station and delta approaches and continues today through use of delta, extreme value, and oxygen–carbon methods (Supplementary Figure S2). Tracer techniques were first applied in the 1980s, comprising about 10% of studies, increasing in the 1990s, decreasing in the 2000s-2010s and rebounding to 22% in the 2020s with the emergence of new field methods (e.g., argon; Hall and Madinger, 2018), while empirical equations plateaued during the 1990s at 23% and have remained relatively stable since. Interface flux and modeling approaches only became common after 2010 (10–15%), mainly for benthic and hyporheic studies. Overall, the methods for GPP, ER, and gas exchange have remained stable since the 1980s, comprised by in-stream DO and tracers (although computational methods have evolved), while longer-term methodological shifts reflect a decline in chamber-based studies and increasing diversification in approaches used to quantify benthic and hyporheic exchanges.

4.2 Which oxygen exchanges are most frequently reported and co-reported?

Reporting effort was highly variable across exchange types (Figure 5a). Whole-stream GPP and ER dominated, with each reported in 33% of studies, followed by surface-atmosphere exchange (18%), together comprising the essential elements of Odum's (1956) whole-stream metabolism framework and accounting for 84% of all reported measurements. Reporting of underlying process domains was comparatively limited, with benthic exchanges appearing in 7% of studies, SOD in 5%, water column in 4%, and hyporheic respiration in less than 1%.

Co-reported exchanges further illustrate this pattern. GPP and ER were jointly reported in 65% of studies, and gas exchange was commonly paired with one or the other (31% with ER; 30% with GPP; Supplementary Figure S3). Beyond whole-stream and surface-atmosphere exchanges, co-reporting declined sharply: benthic production and benthic respiration were the next most frequently paired exchanges (7%), followed by water column production with gas exchange (5%), and water column production with water column respiration (4%). Notably, hyporheic respiration and SOD were never reported together, and when they were measured, they were typically paired alongside benthic or water column exchanges. The distribution of unique exchanges per study further emphasizes this imbalance, where 94% of studies reported three or fewer oxygen exchanges, indicating that most studies focus on whole-stream metabolism, often paired with gas exchange, while omitting the water column, benthic, and hyporheic and sediment measurements contributing to whole-stream oxygen signals (Figure 5b).

4.3 How variable are the oxygen exchanges in streams and rivers?

Among all oxygen exchanges standardized to fluxes (except gas exchange, which is reported as k₆₀₀), site-level mean magnitude varied among groupings (Figure 6; Supplementary Table S2). Within whole-stream metabolism, ER exceeded GPP on average (means = 8.0 and 4.4 g O₂ m⁻² day⁻¹; n = 364 and 371), meaning most streams were heterotrophic (consistent with Bernhardt et al., 2022). For water column observations, production was higher than respiration on average (means = 2.6 and 1.0 g O₂ m⁻² day⁻¹; n = 19 and 22). In contrast, benthic respiration (mean = 11.4 g O₂ m⁻² day⁻¹; range = 3.5 × 10⁻⁴ to 70 g O₂ m⁻² day⁻¹; n = 49) was both higher on average and more variable than benthic production (mean = 2.5 g O₂ m⁻² day⁻¹; range = 0.05–9.1 g O₂ m⁻² day⁻¹; n = 25). Hyporheic respiration was markedly higher on average than SOD, though reported less frequently (means = 7.4 and 1.3 g O₂ m⁻² day⁻¹; n = 6 and 54).

Oxygen exchanges differed in their variability, expressed as the standard deviation of log-transformed fluxes (σ ln) to account for the right skew of the data (Figure 5c; Supplementary Table S2). Hyporheic respiration was the most variable (σ ln= 3.9), followed by benthic respiration (σ ln= 2.4), SOD (σ ln= 1.9), and gas exchange (k₆₀₀; σ ln= 1.8), while whole-stream metabolism ( GPP σ ln = 1.5 ; ER σ ln = 1.3 ), benthic production ( σ ln = 1.4 ), and water column fluxes ( production σ ln = 1.0 ; respiration σ ln = 1.1 ) were comparatively less variable. Subsurface exchanges were simultaneously the most variable and the least frequently measured (see section 5.3).

Oxygen exchange magnitudes varied systematically with stream size and land use, consistent with well-established patterns in whole-stream GPP and ER (Vannote et al., 1980; Bernot et al., 2010; Finlay, 2011; Hoellein et al., 2013; Bernhardt et al., 2022), though comparable syntheses of underlying compartments across these gradients are limited. GPP and ER both peaked in mid-sized streams (means = 6.5 and 10 g O₂ m⁻² day⁻¹) relative to small (2.8 and 6.8) and large (5.2 and 5.6), while gas exchange coefficients (k₆₀₀) declined with stream size (mean = 66 d⁻¹ in small streams and 7.0 d⁻¹ in mid and large). Benthic production was relatively consistent across stream size (means = 2.3–2.7 g O₂ m⁻² day⁻¹), whereas benthic respiration peaked in mid-sized streams (mean = 15 g O₂ m⁻² day⁻¹), and SOD remained low across all sizes (means ranging from 0.95 to 1.6 g O₂ m⁻² day⁻¹); limited water column and hyporheic data constrained inference across stream size. Across land use, GPP and ER were highest in agricultural streams (means = 7.1 and 10.5 g O₂ m⁻² day⁻¹) and lowest in undisturbed (means = 2.8 and 5.8 g O₂ m⁻² day⁻¹), with all categories heterotrophic. Benthic respiration showed the strongest land use signal, peaking markedly in urban streams (mean = 23.6 g O₂ m⁻² day⁻¹) and declining through mixed (7.9), agricultural (2.8), and undisturbed systems (1.1), while k₆₀₀ was highest in undisturbed streams (127.9 d⁻¹) and substantially lower in agricultural (16.4 d⁻¹), mixed (8.2 d⁻¹), and urban (5.4 d⁻¹). Together, these patterns are consistent with undisturbed streams favoring subsurface metabolism in headwaters and increasing water column activity downstream, while elevated benthic respiration in urban and mid-sized streams suggests human modification can disrupt this expected metabolic hierarchy (Ledford et al., 2021; Supplementary Figure S4, Table S3).

Beyond overall magnitudes, we identified 11 studies reporting concurrent process domain and whole-stream measurements (GPP: n = 26 sites; ER: n = 24 sites) that allow us to assess how individual compartments contribute to whole-stream oxygen budgets. We used stream size as an organizing principle because it provides a common basis for comparing observed patterns with ecological theories of longitudinal change in river networks (e.g., Vannote et al., 1980), while recognizing that land use and other disturbances may alter these expected patterns. Given the small sample sizes and skewed distributions within each stream-size category (n = 3–13 sites), we used medians to summarize process-domain fractions and compare their relative contributions across stream sizes (Figure 7). In small streams, hyporheic respiration was the largest ER contributor (median = 63%; range = 43–80%), followed by water column (median = 21%), benthic (median = 18%), and SOD (median = 6%), while water column production dominated GPP (median = 93%) despite benthic production spanning the widest range of any process domain (7–182%). In mid-sized streams, benthic respiration dominated ER (median = 73%) with no hyporheic measurements available, while water column and benthic production contributed comparably to GPP (medians = 21 and 36%). Large streams showed the most constrained patterns, with water column respiration approaching whole-stream ER (median = 97%) and benthic production contributing minimally to GPP (median = 7%). The framing of process domain fluxes relative to whole-stream metabolism is explored further in sections 4.4, 5.2, and 5.3.

4.4 What complementary information and metadata are we reporting alongside oxygen exchange data?

Oxygen dynamics in streams are shaped by complex interactions among physical, chemical, and biological factors, with dominant controls expected to vary as we consider different exchanges and settings (Figure 8; Supplementary Figure S5). Despite this heterogeneity, some variables consistently emerge as more commonly reported regardless of which oxygen exchange(s) were included in a study. Discharge and temperature were the most common covariates, reported in 70 and 56% of all studies, respectively. Among total, particulate, and dissolved nutrient concentrations nitrogen was most often reported (47% of studies), followed by phosphorus (37%), and carbon (26%). Light-related measurements and standing stock indicators were included in about 30% of studies, while turbidity (13%), alkalinity (10%), and canopy cover (6%), were less frequently reported.

Among surface water variables, temperature, nitrogen, and light were most commonly reported, particularly for whole-stream, water column, and benthic exchanges, aligning with expectations from decades of metabolism literature (Bernhardt et al., 2022). Light was reported in 77% of water column production studies and 59% of benthic production. Standing stocks were also frequently reported in water column, benthic and whole-stream metabolism studies (36–50%), but less often reported with gas exchange (26%), SOD (19%), and hyporheic respiration (8%).

By contrast, subsurface variables were rarely reported across all groupings. Groundwater DO—which is necessary to correct metabolism estimates for groundwater inputs—had highest reporting for hyporheic respiration (50%) and SOD (6%), followed by whole-stream (3% for GPP and ER) and surface-atmosphere (1%), and was not reported at all for water column and benthic exchanges. Hydrogeologic flow properties, which describe the ease with which water moves through subsurface materials, such as hydraulic conductivity and transmissivity, were reported in about 15% of whole-stream metabolism, 12% of SOD, and 25% of hyporheic respiration studies. Hydrogeologic transport metrics characterizing subsurface water fluxes and timescales such as residence time, hyporheic exchange fluxes, dispersivity and diffusion coefficients were included in fewer than 13% of studies, though they were highest in hyporheic studies (50%). Organic and inorganic species of carbon were the least reported nutrients across all studies, but most often included with measurements of the water column, benthos, and hyporheic zone, reflecting the coupling between organic carbon consumption and inorganic carbon production during heterotrophic metabolism.

While the influence of the subsurface on in-stream processes is widely acknowledged, our synthesis demonstrates that subsurface variables remain the least reported when studying DO dynamics in flowing waters. This gap is not a result of oversight but rather reflects methodological challenges of measuring processes that are spatially heterogeneous, transient, and operate across several orders of magnitude (Strayer et al., 2003). Variables describing subsurface transport and transformation are inherently site-specific; hyporheic fluxes, for example, are generally larger and easier to measure in steep headwater streams than in low-gradient rivers where exchange rates are slower and residence times longer (González-Pinzón et al., 2015). Similarly, carbon pools encompass a wide range of forms (e.g., organic vs. inorganic, dissolved vs. particulate), and the most relevant reporting metric depends on the dominant pathway of heterotrophic metabolism and the research objective.

The heterogenous reporting of metadata across studies is not unique to stream oxygen exchanges but reflects a broader challenge shared across interdisciplinary fields. Understanding oxygen dynamics requires moving beyond single-domain perspectives and instead considering multiple, interacting controls that operate simultaneously (Wohl, 2017; Wymore et al., 2023). Adopting a common minimum reporting framework that includes hydrologic context, flow regime, surface water chemistry, and subsurface characteristics alongside oxygen exchange measurements would enable future studies to resolve how these interacting physical, chemical, and biological controls vary across stream types and conditions.

5 Current challenges and opportunities to advance our understanding of lotic oxygen dynamics

5.1 Interpreting oxygen exchange magnitudes in the context of a whole-stream oxygen budget

Our synthesis of 230 empirical studies provides a quantitative basis for comparing the magnitudes and variability of oxygen exchanges and their contributions to whole-stream oxygen budgets. Across exchange types, ER consistently exceeded GPP (means = 8.0 and 4.4 g O₂ m⁻² d⁻¹, respectively), confirming the prevalence of heterotrophy in lotic ecosystems (Bernhardt et al., 2022). Yet the magnitudes of underlying process domains revealed a more complex picture and remaining knowledge gaps. Consistent with ecological theory (Vannote et al., 1980), water column exchanges exhibited lower variability and narrower ranges than benthic and subsurface exchanges, whereas hyporheic respiration showed the greatest variability despite being among the least frequently measured exchanges. Together, these patterns suggest that the most spatially heterogeneous and potentially dynamic oxygen-consuming pathways remain among the least resolved components of stream metabolism.

Despite decades of advancement in estimating GPP, ER, and k₆₀₀, relatively few studies quantify the water column, benthic, hyporheic, and sediment pathways contributing to those integrated signals simultaneously. Among the 230 studies synthesized, only 11 (5%) reported concurrent process domain and whole-stream measurements. Within this limited dataset, the dominant process domain shifted systematically with stream size, as hyporheic respiration contributed most to ER in small streams, while water column respiration approached whole-stream ER in large rivers. Similarly, water column production dominated GPP estimates across most stream sizes, except for mid-sized streams, where benthic production contributed substantially to whole-stream fluxes. These patterns broadly align with ecological expectations along the river continuum but do not account for differences in land use or other disturbances, as small streams were primarily agricultural, mid-sized streams were predominantly mixed-use, and large rivers represented agricultural, mixed-use, and undisturbed settings. Although geographically diverse, spanning North America, South America, Australia, and New Zealand, this sparse dataset underscores the need for broader concurrent measurements of oxygen exchange across stream sizes and land-use settings to better resolve the compartment-level processes underlying whole-stream metabolism and inform process-based modeling frameworks.

The limited number of studies measuring concurrent fluxes nonetheless demonstrates the value of resolving multiple oxygen exchanges within the same stream. Studies that quantified multiple oxygen and carbon exchanges simultaneously—including Acuña et al. (2011), Hogsett and Goel (2013), and Oliver and Merrick (2006), which each reported eight exchanges (excluding SOD), and Chessman (1985), which reported six—gained important information on the spatial and temporal variability of different exchanges, with controls shifting across hydrologic conditions, disturbance, and position along the river continuum. Key insights included the importance of specific processes and compartments, such as the hyporheic zone accounting for a majority of ER and floating macroalgae dominating GPP (Acuña et al., 2011), as well as identifying streams where net water column production exceeded net benthic production (Oliver and Merrick, 2006). Importantly, ecosystems with different metabolic pathways can exhibit similar whole-stream signals, including similar GPP: ER ratios. For example, Oliver and Merrick (2006) showed that sites with different dominant carbon sources (e.g., phytoplankton vs. benthic) can have similar NEP values, underscoring the importance of knowing how different compartments are contributing to whole-stream productivity and food web energetics.

5.2 Methods and scaling influence what we can resolve in the oxygen budget

Our synthesis shows that methods have evolved unevenly across oxygen exchanges. Whole-stream GPP and ER have shifted from discrete diel DO measurements toward continuous sensor-based approaches, while the underlying metabolism equations have diversified from primarily single-station models to include two-station, delta, extreme value, and coupled oxygen–carbon approaches. Gas exchange methods have remained comparatively consistent, relying primarily on in-stream DO approaches, tracers, empirical hydraulic equations, and more recently, Bayesian parameter estimation, where tracer selection and normalization toward k₆₀₀ have improved comparability. In contrast, benthic exchange methods have begun to diversify beyond traditional chambers, particularly through interface-flux approaches such as aquatic eddy covariance and sensor-equipped floating probes, which may offer a promising path for resolving in situ benthic oxygen dynamics in larger rivers. However, aquatic eddy covariance is poorly suited to smaller, shallower streams, where the water column and benthos are difficult to distinguish, and where separating hyporheic and groundwater influence becomes particularly challenging. SOD remains largely chamber-based, especially through laboratory incubations, whereas hyporheic studies have increasingly incorporated process-based and reactive transport models that emphasize residence time, reactive transport, and exchange fluxes.

Despite the evolution of some oxygen exchange methods, compartments are typically studied in isolation using methods optimized for specific spatial scales, reinforcing methodological and disciplinary silos that hinder integration. These approaches range from patch-scale chambers and incubations that quantify sediment and benthic exchanges to sub-reach water column metabolism assays and reach-scale measurements of GPP, ER, and gas exchange (Box 1). This fragmentation is compounded by inconsistent reporting units: gas exchange, for example, is typically reported as a rate constant (d⁻¹) or as a velocity (m d⁻¹), but the literature contains at least 26 distinct gas exchange coefficients (Supplementary Figure S1), although recent years have seen a convergence toward using a normalized k₆₀₀ (d⁻¹). Whole-stream metabolism is typically expressed in areal units (g O₂ m⁻² d⁻¹), whereas water column, benthic, and hyporheic studies are often reported as volumetric rates (g O₂ L⁻¹ d⁻¹), with some hyporheic and reactive transport studies additionally expressing respiration as first-order DO uptake or decay coefficients (d⁻¹), further complicating direct comparison across studies and exchange types without appropriate metadata.

Although some efforts have attempted to bridge methodological and scaling challenges—for example, by comparing patch-scale and whole-stream metabolism estimates (e.g., Tromboni et al., 2017) or by linking metabolism and nutrient uptake dynamics (e.g., Hoellein et al., 2009)—these integrative approaches remain rare in the literature. The integration of individual fluxes, such as gas exchange, has already been shown to be extremely valuable for constraining open-channel metabolism methods. We submit that similar improvements in the accuracy of and controls on whole-stream metabolism will emerge from routinely connecting water column and subsurface fluxes to common DO concentration inferences.

Translating process domain rates to whole-stream fluxes requires scaling information that varies in complexity with the compartment (see section 4.4 and Figure 8). Scaling water column rates to whole-stream fluxes requires estimates of integrated channel depth, which is generally straightforward, though varies in space and time (e.g., Oliver and Merrick, 2006). Benthic fluxes require spatial extrapolation across a heterogeneous streambed, where substrate type, organic matter, and disturbance history can create variation at patch scales (Fellows et al., 2006; Lee-Cullin et al., 2018). Gas exchange scaling depends on stream size, with morphology and turbulence dominating in small streams (Raymond et al., 2012) and wind and depth gradients in larger rivers (Hall and Ulseth, 2020). Reach-scale hyporheic fluxes may be the most complex to quantify, requiring not just a respiration rate but the exchange flux, active sediment volume, and residence time (Gomez-Velez and Harvey, 2014), the subsurface metadata least frequently reported across all exchange types (Figure 8). Without these, even concurrent compartment measurements cannot be meaningfully summed or compared to whole-stream signals, particularly for subsurface exchanges where process boundaries remain poorly defined.

5.3 The hyporheic zone as an under-resolved component of the oxygen budget

Hyporheic respiration is simultaneously the most variable (σ_ln = 3.9) and least frequently reported exchange, appearing in less than 1% of studies (Figures 5a,c). This pattern likely reflects both methodological heterogeneity and a disciplinary disconnect between hyporheic and stream metabolism research. Among the 11 studies identified, hyporheic respiration was quantified using a diverse suite of approaches, including in-stream DO methods (36%), field chambers (27%), modeling (18%), tracers (9%), and other approaches (9%), with no dominant methodological framework. These methods integrate respiration across different spatial and temporal scales and therefore produce estimates that are not directly comparable. Moreover, many investigations of hyporheic oxygen consumption are conducted within groundwater hydrology, biogeochemistry, or reactive transport studies rather than whole-stream DO or metabolism research and were therefore not captured in this synthesis.

Terminology and methodological overlap further complicate synthesis of stream and river DO fluxes. Sediment oxygen demand (SOD), total oxygen demand (TOD), benthic chamber measurements, and hyporheic respiration resolve overlapping processes through distinct approaches with different spatial footprints and temporal assumptions. Hogsett and Goel (2013) defined SOD as the total abiotic and biotic oxygen consumption at the sediment surface including hyporheic exchange, while TOD excludes hyporheic exchange. This distinction matters because much of the broader SOD literature relies on ex situ or in situ sediment cores that capture demand at a fixed depth without accounting for advective exchange. Thus, reported SOD values may reflect anything from surface-only demand to an unknown mixture of benthic and hyporheic contributions depending on chamber design and study site. Coenen et al. (2019) emphasized the importance of standardizing SOD measurements to enhance cross-study comparability, which we extend to all subsurface oxygen exchanges, using SOD in this definition and reserving hyporheic respiration for measurements that account for advective exchange.

Methodological and scaling challenges also complicate the application of the Odum (1956) framework, whose assumptions are challenged in heterogeneous streams where subsurface processes influence DO dynamics. We caution against differencing whole-stream metabolism estimates with independently measured benthic and water column exchanges, as this approach can introduce statistical uncertainty and process error in streams where hyporheic respiration contributes substantially to DO dynamics and reach-scale GPP and ER reflect partially unresolved subsurface processes (Naegeli and Uehlinger, 1997; Gomez-Velez et al., 2026). Rather than treating methods as interchangeable, complementary approaches should be combined to better constrain which processes are contributing to whole-stream metabolism. Modified two-station metabolism models, for example, are required to account for groundwater influences on oxygen dynamics that may bias estimates of GPP and ER (Hall and Tank, 2005). Yet, the use of two-station models (with or without groundwater corrections) has declined from a peak in the 1990s when ~50% of whole-stream metabolism studies employed a two-station design to ~25% in the 2000s and 2010s. Re-diversifying beyond single-station models is particularly important for systems that violate the core assumptions underlying those approaches. Development of modern analytical toolkits for two-station inferences like those available for single-station may reverse this trend, with recent advances in variable-flow systems offering a promising template (Payn et al., 2017; Bishop et al., 2026).

5.4 Building toward an integrated stream oxygen budget

Resolving oxygen exchanges across compartments is increasingly urgent as environmental change reshapes the processes governing stream oxygen dynamics. Warming temperatures accelerate respiration while reducing DO concentrations (Demars et al., 2011; Zhi et al., 2023), hydrologic fragmentation and expanding intermittency restructure subsurface connectivity and metabolic regimes (Datry et al., 2014; Costigan et al., 2015; Messager et al., 2021; Nelson et al., 2023), altered groundwater recharge changes the balance between surface and subsurface oxygen demand (Gleeson et al., 2012; Famiglietti, 2014), and eutrophication amplifies water column and benthic respiration in ways that whole-stream metabolism cannot resolve alone (Dodds and Cole, 2007; Finlay, 2011; Jankowski et al., 2021).

Climate-driven hydrologic variation will particularly impact intermittent streams (Ward et al., 2020), which are disproportionately important for biogeochemical processing (Datry et al., 2014; Stegen et al., 2025) and comprise roughly half of global rivers (Messager et al., 2021). Yet only 4% of studies in our synthesis included intermittent streams, with most work instead focused on perennial streams and baseflow conditions, leaving less known about how whole-stream and compartment-specific respiration respond to changing discharge and hydrologic connectivity. This matters because repeated drying and rewetting cycles introduce temporal heterogeneity that can amplify, delay, mask, or decouple respiration signals from hydrologic connectivity (Ward et al., 2019; Arce et al., 2021; Bretz et al., 2023), underscoring a broader limitation in interpreting diel DO under unsteady flow. Coordinated initiatives including the Dry Rivers Research Coordination Network (Datry et al., 2023; Price et al., 2024) and the Aquatic Intermittency Effects on Microbiomes in Streams (National Science Foundation, 2025) project are beginning to address these gaps through cross-biome comparisons and the development of methods appropriate for non-perennial streams.

Beyond hydrologic connectivity, building toward an integrated oxygen budget will require coupling DO with other biogeochemical cycles. While whole-stream metabolism provides valuable insight into aerobic carbon cycling (i.e., presuming oxygen is consumed aerobically), it represents an integrated metric that cannot inform all underlying transformations occurring within streams. Integrating DO with complementary measures such as other naturally occurring dissolved gases (e.g., CO₂, CH₄, Ar), nutrient spiraling, and study of individual fluxes (e.g., water column, benthic, and hyporheic zones) can directly link oxygen dynamics with nutrient and carbon transformations. For example, by coupling DO-derived metabolism with organic carbon and gas exchange, Hall et al. (2016) showed how oxygen dynamics can be used to infer large-scale patterns in carbon metabolism and transport across river networks. Similar cross-scale measurement approaches have illustrated how DO dynamics relate to nitrification, denitrification, and DOC metabolism across land use gradients (Hanrahan et al., 2018; Plont et al., 2022), collectively revealing that oxygen analyses offer enhanced insights into ecosystem processes when considering their tight coupling to carbon and nutrient transformations.

Moving forward, advancing multi-solute and cross-disciplinary frameworks will be critical to link DO-based metabolism with CO₂, nutrient, and redox dynamics. Discrepancies between CO₂- and O₂-based estimates of stream metabolism (Bernal et al., 2022; Shangguan et al., 2024) highlight how metabolic processes exist that are not captured in DO measurements alone as well as the different spatial scales associated with different gas dynamics. Decoupling may arise from anaerobic respiration that does not impact DO concentration, or hyporheic pathways that induce long lags in water column DO signals (Gomez-Velez et al., 2026). Sources of error are also compounded by the widespread use of a 1:1 O₂–CO₂ stoichiometric ratio when converting DO-based metabolism to carbon units as this ratio is known to vary (Trentman et al., 2023), particularly in ecosystems with strong groundwater influence (Rocher-Ros et al., 2025). Together, these patterns underscore that aerobic and anaerobic respiration reflect distinct redox pathways that shape organic matter decomposition and nutrient turnover, yet quantifying anaerobic respiration remains challenging due to its spatial heterogeneity. Studies such as Son et al. (2022) demonstrate that hyporheic exchange can drive large-scale variability in anaerobic metabolism, underscoring the need to explicitly represent subsurface pathways in metabolic frameworks. Developing modeling approaches capable of partitioning autotrophic, heterotrophic, and anaerobic signals (Hall and Beaulieu, 2013; Bertuzzo et al., 2022; Diamond and Bertuzzo, 2025) will be key to linking empirical data and our process-based understanding.

6 Conclusion

Advancing dissolved oxygen science requires moving beyond single-flux measurements toward the concurrent quantification of oxygen exchanges across process domains that collectively give rise to whole-stream signals. Evidence from the relatively few studies that have quantified multiple oxygen exchanges demonstrates that this added complexity yields disproportionate insight. These studies reveal strong spatial and temporal variability in process domains, shifts in the relative dominance of individual exchanges (e.g., hyporheic respiration or water column production), and cases where fundamentally different underlying pathways produce similar whole-stream fluxes. Importantly, even partial resolution of multiple exchanges has been shown to improve process attribution by identifying hyporheic exchange as a hidden control on sub-daily oxygen dynamics, explaining apparent variability and bias in GPP, ER, and gas exchange estimates that would otherwise be attributed to noise or equifinality. As a result, whole-stream estimates of GPP, ER, or NEP alone cannot resolve where oxygen and carbon are primarily produced or consumed. Distinguishing whether oxygen dynamics are dominated by water column, benthic, or hyporheic and sediment processes has direct implications for the supply and demand of energy in food webs, aquatic ecosystem health, management decisions, and the interpretation of stream DO signals from which metabolic fluxes are inferred.

At the same time, boundaries between compartments are often diffuse and temporally dynamic, further complicating attribution. Sediments may integrate benthic and hyporheic signals, while under certain hydrologic conditions subsurface flowpaths can dominate metabolic activity, influencing in-stream diel DO concentrations. Although oxygen dynamics retain distinct process signatures, their effective integration length makes them sensitive to small-scale changes; for instance, more turbulence and mixing can reset oxygen signals and either amplify or dampen hyporheic contributions depending on flow and season. These sensitivities are especially pronounced in intermittent streams, where dynamic subsurface flowpaths contribute to streamflow yet remain underrepresented in the literature. While our review highlights the need for compartment-based measurements, we also encourage a shift toward a more integrated perspective, in which oxygen production, consumption, and transport are quantified along hydrologic pathways spanning the water column, benthos, hyporheic zone, sediments, and groundwater. Such an approach better reflects how oxygen moves through and is processed within stream networks, with hydrologic connectivity acting as a master variable that mediates fluxes among compartments and regulates when and where specific exchanges may dominate.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frwa.2026.1954977/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Mohammad Safeeq, University of California, Merced, United States

Reviewed by: Yang Yang, Auburn University, United States

Piotr Kłosiński, Nicolaus Copernicus University in Toruń, Poland