High rates of daytime river metabolism are an underestimated component of carbon cycling

The study addresses whether traditional river metabolism estimates based solely on diel dissolved oxygen (O₂) concentrations underestimate gross primary production (GPP) and ecosystem respiration (ER) by assuming ER is constant over day and night. The authors investigate if incorporating stable oxygen isotope signatures of dissolved oxygen (δ¹⁸O₂) alongside O₂ concentrations reveals higher daytime ER and GPP and faster carbon cycling. This question is important because accurate quantification of aquatic carbon fixation and respiration underpins understanding of riverine contributions to the global carbon cycle and informs how local processes scale to global CO₂ fluxes.

Traditional open-channel metabolism models use diel O₂ concentration dynamics to estimate GPP and ER (e.g., Odum 1956; Holtgrieve et al. 2010), typically assuming constant ER over 24 h. Stable isotope approaches using δ¹⁸O₂ have been developed to better constrain metabolism and reaeration (e.g., Tobias et al. 2007; Hotchkiss & Hall 2014) and indicate potential daytime increases in ER, but prior applications were limited (e.g., to a few temperate streams). Background work documents diel isotopic dynamics of O₂, influences of temperature and hydrodynamics on isotope fractionation, and the roles of primary production-derived DOC and photochemical processes in fueling heterotrophic metabolism. This study expands these approaches across multiple biomes and ecoregions to evaluate generality.

• Study design and sites: Fourteen 24-h experiments were conducted during 2017–2018 across temperate (mountain steppe, desert), tropical, and subarctic rivers spanning a gradient of productivity, hydrology, and land use (Table 1). Most sites were minimally impacted; Mekong and Sekong had higher impacted land cover.
• In situ measurements: Dissolved O₂ concentration and temperature were logged every 10 min for at least 24 h using MiniDOT sensors, with pre/post calibrations in air-saturated water and equilibration prior to data use. Photosynthetically active radiation (PAR/PPFD) was logged every 10 min with calibrated sensors.
• Isotope sampling: δ¹⁸O₂ grab samples were collected every 2 h during the same 24-h period in pre-evacuated vials preserved with HgCl₂. δ¹⁸O₂ was analyzed via IRMS (Micromass Isoprime with EA-GC), and water δ¹⁸O via Picarro L2130-i. Analytical uncertainties: ±0.2‰ for δ¹⁸O₂ and ±0.1‰ for water-δ¹⁸O.
• Site characterization: Conductivity, slope, depth, and velocity were measured across transects; biofilm ash-free dry mass (AFDM) was measured from 8–12 rocks per site.
• Respiration isotope fractionation (αr): Sealed recirculating Plexiglas chambers with site substrata (rocks, sediments, macrophytes as appropriate) were used to estimate respiration fractionation factors (αr) concurrently with diel sampling, following prior methods. αr was computed from pre/post incubation δ¹⁸O₂ and O₂ concentrations.
• O₂-only metabolism model: Open-channel single-station Bayesian inverse modeling estimated daily GPPO₂ and ER₀₂ from Eq. (3), using fixed site-specific reaeration (K₀) derived from prior K₆₀₀ estimates (BASE) with Schmidt number and temperature corrections. Parameters were inferred by fitting modeled to observed O₂ time series.
• Coupled O₂–δ¹⁸O₂ model: An updated diel isotope model coupled high-frequency O₂ with 2-hourly δ¹⁸O₂ to estimate daily δ¹⁸O₂-GPP and δ¹⁸O₂-ER via Eqs. (4a, 4b). The model introduced a diel metabolism term (dielMET) that adjusts GPPO₂ and ER₀₂ equally to capture diel variation inferred from δ¹⁸O₂ dynamics, incorporating isotope fractionations for gas exchange (α), respiration (αr from chambers), and photosynthesis (αp), with atomic fractions of ¹⁸O in water, air, and dissolved O₂.
• Bayesian inference: Posterior distributions for ER₀₂, GPPO₂, and dielMET were estimated via MCMC (R mcmc/metrop) with at least 200,000 iterations, removal of 10,000 burn-in, and nominally informative priors for GPPO₂ and ER₀₂ (from O₂-only fits) and uniform priors for dielMET (0–100 g O₂ m⁻² d⁻¹). For each site, models using minimum, mean, and maximum measured αr were compared; the best fit (lowest sum of squared differences for O₂ and δ¹⁸O₂) was selected.
• Temperature normalization: To isolate temperature effects, metabolism rates were normalized to 20 °C using a published relationship, yielding 20δ¹⁸O₂-ER and 20δ¹⁸O₂-GPP for comparison with O₂-only estimates.
• Statistics: Multiple linear regression with stepwise AIC selection examined predictors of diel 20δ¹⁸O₂-ER magnitude among sites, testing variables including impacted land use (%), 20δ¹⁸O₂-GPP, conductivity, AFDM, slope, depth, and velocity; ANOVA assessed variable contributions.

• Across all 14 rivers spanning tropical, temperate (mountain steppe, desert), and subarctic biomes, δ¹⁸O₂-based models indicated higher daytime ER than nighttime ER. The maximum diel change in ER was 113 g O₂ m⁻² d⁻¹ and the minimum was 1 g O₂ m⁻² d⁻¹.
• Temperature-normalized (20 °C) ER and GPP derived from δ¹⁸O₂ data were substantially higher than O₂-only estimates: ER was 1.1–87× higher and GPP was 1.5–22× higher.
• Differences between δ¹⁸O₂-derived and O₂-only estimates were especially large in some biomes, reaching up to roughly 1700% (tropics) and 2200% (desert rivers in Mongolia) for ER.
• A positive relationship between daytime δ¹⁸O₂-ER and δ¹⁸O₂-GPP suggests that photosynthetic activity during the day stimulates heterotrophic respiration, likely via DOC exudation and/or photochemical DOC processing.
• Normalizing to 20 °C did not erase the pattern of higher daytime ER, indicating temperature alone does not explain elevated daytime ER; patterns were evident even in tropical rivers with minimal diel temperature variation.
• δ¹⁸O₂-ER and δ¹⁸O₂-GPP estimates remained significantly higher than ER₀₂ and GPP₀₂ across biomes despite model uncertainties, indicating traditional O₂-only approaches underestimate riverine carbon cycling rates.

The findings directly address the hypothesis that traditional diel O₂-based metabolism models underestimate river metabolism by assuming constant ER over 24 h. By coupling O₂ concentrations with δ¹⁸O₂ dynamics and allowing diel variation (dielMET), the study consistently revealed elevated daytime ER alongside higher GPP across all sites and biomes. Temperature normalization demonstrated that increased daytime ER is not primarily a temperature artifact. Mechanistically, the positive association between δ¹⁸O₂-GPP and δ¹⁸O₂-ER supports the idea that daytime photosynthesis enhances heterotrophic activity through rapid use of photosynthate-derived DOC and/or photolysis-increased DOC availability. Photorespiration was considered but is unlikely to be the dominant driver given observed timing relative to light peaks, biofilm vertical structure and light attenuation, and the fact not all primary producers exhibit significant photorespiration. These results imply faster within-day cycling of oxygen and carbon and point to a more active microbial loop and benthic biofilm processes than inferred from O₂-only models. The work has implications for understanding drivers of CO₂ fluxes, denitrification, methanogenesis, and food web support in rivers, and suggests broader relevance to other benthic biofilm-dominated ecosystems.

Ignoring day–night differences in ecosystem metabolism leads to substantial underestimation of riverine GPP and ER. Coupled O₂–δ¹⁸O₂ Bayesian modeling across diverse rivers showed consistently higher daytime ER and higher overall metabolism than O₂-only models, indicating more rapid microbial carbon cycling and greater importance of primary production to river food webs than previously recognized. These insights challenge prevailing assumptions in metabolism modeling and have implications for interpreting CO₂ generation and extrapolating from local to global scales. Future work should develop high-frequency δ¹⁸O₂ measurement technologies, integrate simultaneous DIC concentration and δ¹³C measurements, and perform pulse–chase isotope experiments to resolve short-timescale carbon fixation, exudation, heterotrophic uptake, and fates under changing flow regimes, eutrophication, and climate.

• Measurement frequency: δ¹⁸O₂ data were collected every 2 h (grab sampling), yielding fewer observations than 10-min O₂ sensor data and contributing to higher uncertainty in diel isotope model estimates, especially in low-GPP, high-gas-exchange rivers where physical processes can mask biological signals.
• Model parameter uncertainty: Estimates depend on the respiration fractionation factor (αr), which varies among sites and is challenging to quantify at ecosystem scale. Although site-specific chamber-derived ranges were used and best-fit αr selected, residual uncertainty remains.
• Gas exchange and physical masking: High air–water gas exchange rates reduce the isotopic signal-to-noise, increasing uncertainty in ER and GPP estimates from δ¹⁸O₂.
• Short deployment duration: Analyses were based on single 24-h periods per site; multi-day deployments with high-frequency δ¹⁸O₂ would better constrain variability and model fits.
• Mechanistic attribution: While evidence favors DOC exudation/photochemistry stimulating daytime ER, the study did not directly measure exudation or photolysis; photorespiration cannot be entirely ruled out.