Drought alters the biogeochemistry of boreal stream networks

Hydrological droughts—originating from persistent atmospheric anomalies and propagating through the hydrological cycle—are intensifying due to climate change and human activities. Although well studied in arid and semiarid regions, their impacts on high-latitude (north of ~55°N) streams, which constitute ~33% of the global river network and drain vast soil carbon reserves, are poorly known. Boreal headwaters play critical roles in processing terrestrial dissolved organic matter (DOM), emitting greenhouse gases (GHGs), supporting biota, and regulating downstream chemistry. Drought can alter stream biogeochemistry by reducing hydrologic transport of substrates, increasing water residence time (WRT), and restricting dissolved oxygen (O2) resupply, thereby shifting metabolism from aerobic to anaerobic pathways. Northern streams may be especially susceptible due to high organic matter at the land–water interface and persistent DOM supply. The study tests the hypothesis that drought shifts metabolic pathways of DOM processing under different redox conditions, evaluates network-scale biogeochemical responses, and considers implications for water quality in northern boreal streams.

Prior work shows drought impacts water quality and biogeochemistry via altered hydrology, substrate supply, and redox constraints (e.g., Mosley 2015; Dahm et al. 2003). Metabolic processes follow a thermodynamic sequence of electron acceptors, with O2 depletion promoting anaerobic pathways including methanogenesis (Champ et al. 1979). Headwater streams at high latitudes are important DOM processors and GHG sources (Rasilo et al. 2016; Hotchkiss et al. 2015) and are influenced by riparian peat-rich soils that can impose reducing conditions (Ledesma et al. 2016; Lupon et al. 2019). Drought can either increase or decrease aerobic metabolism depending on context (Harjung et al. 2018; Hosen et al. 2019; Acuña et al. 2005). Stoichiometric analysis of departures of CO2 and O2 from atmospheric equilibrium (ΔCO2:ΔO2) provides insight into dominance of metabolic pathways, with deviations from a 1:-1 line indicating non-aerobic processes (Torgersen & Branco 2007; Vachon et al. 2020). CH4:CO2 ratios serve as a proxy for methanogenesis while minimizing physical effects on absolute concentrations (Stanley et al. 2016; Campeau & Del Giorgio 2014). Monitoring gaps are largest in small headwater streams that make up most network length (Benstead & Leigh 2012; Bishop et al. 2008).

The study combined: (1) a reach-scale drought manipulation experiment (August 2017) in a 1.4-km boreal headwater stream within the Krycklan Catchment Study (KCS), northern Sweden; (2) network-scale observations during the extreme 2018 natural drought; and (3) analysis of long-term (2010–2018) monitoring data.

• Experimental design (2017): Drought simulated by damming a lake outlet feeding the stream for ~2 weeks (Aug 7–18), with background periods before (Aug 3–7) and after (Aug 24–30). Six 50-m segments were instrumented along the reach. Discharge at the reach ends (flumes at C5 and C6) was measured from 10-min stage and rating curves; discharge along the channel was modeled using a 2-m DEM and upslope-contributing area to estimate hourly discharge and lateral groundwater inputs per 50-m cell. Segment-specific mean depth and width from cross-sections allowed calculation of velocity and WRT (segment length/velocity). Salt releases validated Q and WRT estimates (±10%).
• Experimental hydrologic change: Mean discharge across segments dropped from 12.3 to 1.1 L s−1; mean WRT rose from 28.0 to 223.1 min; local WRT spanned 14.3–1061.3 min due to patchy lateral inflows that weakened over time. Water temperature decreased modestly (12.7 ± 1.8 to 10.4 ± 1.1 °C).
• Sampling during experiment: Low-frequency grab sampling (three times during drought, two during background) of surface and hyporheic waters at segment bottoms for O2, NO3−, SO4 2−, CO2, NH4+, CH4, DOC, pH, temperature, conductivity. Hyporheic samples from PVC wells (25–50 cm depth); near-stream groundwater wells at main inflow zones were also sampled. High-frequency sensors recorded O2 (concentration and % saturation) and temperature at 10-min intervals in surface and hyporheic waters at all segments; dissolved CO2 at 10-min intervals at four segments, using a membrane-covered Vaisala sensor.
• Network-scale monitoring (2017–2018): Ten streams spanning Strahler orders 1–5 (0.04–68.9 km2 catchments) with diverse land cover (forests, mires, lakes) were instrumented. Surface O2 (% saturation, concentration) and temperature were recorded every 10 min (miniDOT). CO2, CH4, and DOC were sampled monthly in winter and biweekly in summer–fall. Additional O2 sensors were deployed at six more sites in summer 2017 and 2018; three synoptic chemistry surveys covered 22 headwater streams during summer 2018 drought.
• Long-term datasets (2010–2018): Biweekly to monthly CO2, CH4 and ancillary chemistry at the same ten stations; hourly discharge via permanent H-flumes, converted to specific discharge (mm d−1). Hydrologic conditions classified by historical discharge percentiles: drought (0–10th), low flow (10–20th), baseflow (20–50th), high flow (50–100th).
• Laboratory analyses: pH (Orion 9272 with Ross electrode), DOC (Shimadzu TOC-VCPH after acidification), NH4+ and NO3− (SEAL AutoAnalyzer 3), SO4 2− (ion chromatography), CO2 and CH4 (GC-FID with methanizer; headspace equilibration). DIC species computed from pH and carbonate equilibria; free CO2 >95% of DIC.
• Metabolism modeling: Open-channel single-station method with Bayesian inverse modeling to estimate gross primary production (GPP) and ecosystem respiration (ER); focus on ER (aerobic respiration). Relationships between ER and WRT assessed; drought vs background compared via Wilcoxon Signed-Rank tests.
• Redox and stoichiometric analyses: PCA on O2, CH4, SO4 2−, NO3−, NH4+; dependence of PC1 on WRT via regression. Ratios NH4+:NO3− and CH4:O2 used as indicators of redox shifts and terminal electron-accepting process balance. ΔCO2:ΔO2 stoichiometry (departures from atmospheric equilibrium) evaluated from discrete and high-frequency data; deviations from 1:-1 line indicate non-aerobic processes. Nonparametric percentile regressions used to relate O2 saturation to stream order. CH4:CO2 ratios examined seasonally and across flow classes, using Wilcoxon tests and Loess fits. All statistics performed in R (stats, nlme, vegan, quantreg.nonpar); PCA in XLSTAT; significance at p<0.05.

• Experimental drought (2017) rapidly reduced O2 in surface and especially hyporheic waters; hyporheic O2 fell below 25% saturation once WRT exceeded ~200 min, with a nonlinear decrease in O2 vs WRT.
• Aerobic respiration (ER) decreased significantly during experimental drought from −403 ± 172 to −130 ± 81 mmol O2 m−2 d−1; ER declined nonlinearly with WRT (r2 = 0.41; p < 0.001; n = 111).
• Redox-sensitive indicators shifted toward reduced forms: NH4+:NO3− increased with WRT, implying constrained nitrification and enhanced denitrification; CH4:O2 increased with WRT, indicating progression to anaerobic terminal electron-accepting processes including methanogenesis. Early in drought, lateral groundwater inputs influenced ratios, but signals converged at high WRT (>~1000 min) as lateral inflows waned.
• ΔCO2:ΔO2 analysis showed departures from the 1:-1 aerobic respiration line during drought in both discrete and high-frequency data, with increased centroid and slopes and greater dispersion, indicating persistent CO2 production via anaerobic pathways and diversification of metabolic processes.
• Network-scale 2018 drought effects: Widespread declines in surface-water O2 saturation across 16 monitored sites, most pronounced in headwaters. All 9 headwater streams had at least one O2 excursion <50% saturation; 5 streams had excursions <25%; one was anoxic for much of the summer. Streams draining <2 km2 (65–80% of total network length in KCS) showed potentially severe O2 stress.
• Methanogenesis proxy: CH4:CO2 ratios in 2018 increased relative to 2017 (median 0.014 vs 0.0054), indicating enhanced methanogenesis during extreme low flows. Long-term (2010–2018) data showed CH4:CO2 ratios increased and became more dispersed during low-flow and drought conditions across headwaters.
• Gas fluxes: Despite reduced reaeration (k600) at low flows, CO2 flux varied only modestly with discharge (suggesting evasion constrained by low turbulence), while CH4 fluxes remained stable across the discharge range due to elevated concentration gradients during drought, implying sufficient in-stream CH4 supply to overcome reduced turbulence. Given CH4’s ~30-fold higher global warming potential than CO2, CH4 becomes an emergent component of GHG budgets during low-flow periods.
• Landscape context: Variability in drought severity and duration among sites was mediated by groundwater–stream connectivity, land cover (forest volume, mire cover), topography, soil depth, and presence of lakes. Small forested till catchments were most vulnerable; mires can buffer hydrology but supply reduced solutes; lakes act as storage and can dampen or exacerbate downstream effects.

Findings support the hypothesis that drought shifts boreal stream metabolism from aerobic to anaerobic pathways by increasing WRT and limiting O2 resupply, resulting in accumulation of reduced solutes (NH4+, CH4) and stoichiometric imbalances (ΔCO2:ΔO2). These changes occurred rapidly in the experimental reach and were widespread during the 2018 natural drought, especially in headwaters that dominate network length and ecosystem functioning. The emergence of hypoxia/anoxia indicates potential for water quality degradation and biological stress in small northern streams. Elevated CH4:CO2 ratios and sustained CH4 fluxes under low turbulence suggest that drought can enhance methanogenesis and CH4 evasion from streams, potentially increasing the catchment-scale contribution of aquatic CH4 during drought when adjacent terrestrial systems act more as CH4 sinks. Spatial heterogeneity in responses highlights the role of groundwater connectivity, catchment characteristics, and landscape components (mires, lakes) in propagating or buffering drought effects across networks. Monitoring and managing high-latitude stream health under increasing hydrologic extremes requires attention to small, under-monitored headwaters and to the interactions among physical and biogeochemical controls.

Drought in northern boreal streams induces rapid and pronounced biogeochemical shifts—reduced O2 availability, decreased aerobic respiration, increased reduced solutes, enhanced methanogenesis—and can trigger poor water quality conditions, particularly in headwaters. The 2018 extreme drought revealed network-scale deoxygenation and metabolic transitions that are consistent with patterns observed in a controlled reach-scale manipulation and in long-term records. These results underscore the sensitivity of high-latitude headwater networks to extreme low flows and the potential for increased CH4 significance in GHG budgets during drought. As drought frequency and severity are projected to increase at high latitudes, understanding and predicting stream responses will require resolving landscape controls (groundwater connectivity, land cover, soils, mires, lakes) and improving monitoring of small streams. Future research should quantify whole-stream anaerobic metabolism, disentangle groundwater vs in-stream process contributions during drought onset, assess biological community impacts and recovery, and evaluate how event frequency and duration alter long-term ecosystem functions and services.

• The experimental drought did not simulate terrestrial (on-land) drought; lateral groundwater inflows persisted initially, influencing early-stage chemistry and potentially exaggerating lateral contributions of reduced solutes compared to fully isolated channels.
• Short experimental duration (~2 weeks) limits inference about longer-term adjustments and recovery dynamics.
• Whole-stream anaerobic metabolism was not directly quantified; proxies (ΔCO2:ΔO2, CH4:CO2) were used and have inherent uncertainties and non-biological confounders (e.g., carbonate chemistry effects on CO2).
• Spatial heterogeneity in groundwater–stream connectivity and redox state along the reach and among catchments introduces variability not fully controlled experimentally.
• Gas flux estimates rely on k600 parameterization and may be sensitive to uncertainties in reaeration under very low flows.
• Headwater streams are underrepresented in routine monitoring, potentially limiting broader generalization beyond the KCS without additional replication.