Alkalinity responses to climate warming destabilise the Earth's thermostat

The study addresses how climatic and erosional factors jointly regulate riverine alkalinity generation, a key modulator of the Earth’s carbon cycle. While lithology and discharge are recognized global controls, the distinct roles of temperature and physical erosion remain debated, especially at global scales. Carbonate weathering responds quickly to environmental changes due to rapid dissolution kinetics, making it sensitive to anthropogenic and climatic shifts over sub-millennial timescales. The central research questions are: (1) What are the first-order controls on global riverine alkalinity when accounting for erosion rate? (2) How will projected climate warming alter terrestrial alkalinity fluxes to the ocean? The purpose is to build an empirically grounded, globally applicable function linking normalized alkalinity to physical and climatic predictors and to quantify mid-latitude alkalinity flux changes under future warming scenarios.

Prior global models emphasize river discharge and lithology as dominant alkalinity controls, with acidity supply (mainly from atmospheric/soil CO₂) and physical erosion also exerting first-order influences. The global temperature dependence of carbonate weathering has been debated, but recent syntheses indicate optimal weathering in temperate climates, often described by a Gaussian relation peaking near ~11 °C. Regional studies demonstrate that physical erosion enhances weathering at low-to-moderate rates (supply limitation), though a clear global relationship with alkalinity flux has been elusive. Additional studies highlight that carbonate weathering can be accelerated by increased soil CO₂ from ecosystem respiration and modulated by aridity, while high-relief settings may reduce dissolved loads through CO₂ degassing and secondary CaCO₃ precipitation. The long-term CO₂ balance distinguishes silicate weathering as a net sink beyond CaCO₃ compensation timescales, whereas carbonate weathering is CO₂-neutral over the long term.

Data compilation: The authors compiled 233 catchments worldwide (44°S–51°N) pairing riverine alkalinity with in situ 10Be-derived erosion rates (OCTOPUS database). Alkalinity data were sourced from new measurements (111 sites), GLORICH (76 sites), and additional literature and agencies (46 sites), curated with spatial matching criteria (same river, downstream alignment, ≤25 km separation, limited major tributary influence). Measurements: For selected catchments (2020 campaigns), dissolved inorganic carbon (DIC) and total alkalinity (AT) were measured with standardized instrumentation (Marianda VINDTA 3C, Metrohm 888 Titrando), calibrated against certified reference materials. In situ water temperature, conductivity, and turbidity were recorded. AT and bicarbonate were treated as equivalent for riverine conditions (pH 7–9), verified via CO2SYS calculations (bicarbonate ~98 ± 4% of AT). Normalization and covariates: To remove dilution/evaporation effects, alkalinity concentration was normalized by mean annual runoff (UNH/GRDC composite). Catchment descriptors were derived from OCTOPUS basin outlines: lithology (GLiM), MAT and MAP (WorldClim 2.1), permanent snow/ice and forest cover (GlobCover 2009), soil regolith thickness (DTB depth to bedrock), catchment area (2D), and dams (GOODD). Observations with mean annual runoff <150 mm a⁻¹ were excluded. Modeling: A generalized linear model (GLM) for normalized alkalinity concentration was developed following iterative variable selection, with significant predictors: areal carbonate proportion (combined GLiM classes sc+sm+mtpu), MAT (modeled with a 5th-degree polynomial), natural log of erosion rate, natural log of catchment area, and soil regolith thickness. Runoff was excluded to avoid autocorrelation with the normalized response. A complementary generalized additive model (GAM) for alkalinity flux (alkalinity × runoff) confirmed the significance of the same five predictors. The final GLM (model M5) equation for normalized alkalinity is: exp[−1.163 + 0.01867(carbonate %) − 0.1504(MAT) − 0.009028(MAT)² + 0.005944(MAT)³ − 0.0004681(MAT)⁴ + 0.00001007(MAT)⁵ + 0.2873 ln(erosion) − 0.05615 ln(erosion)² + 0.1342 ln(area) + 0.05078(soil thickness)]. Sensitivity to MAT: Model responses to MAT were explored keeping other predictors at global mean values (carbonate ~22%, erosion 100 mm ka⁻¹, area 1000 km², soil thickness 15 m), with quantitative interpretation limited to MAT 0–20 °C to avoid out-of-sample combinations (e.g., high MAT with high carbonate) and unrealistic fixed erosion in polar regions. Climate impact assessment: For mid-latitude catchments (historical MAT 0–20 °C), historical (1980–2009) and future (2070–2099) MAT fields from ISIMIP (GFDL-ESM4, CMIP6) were used. Runoff and discharge were generated with the HydroPy global hydrology model. To isolate temperature effects, discharge was held constant (historical mean). The GLM yielded historical and future normalized alkalinity, converted to concentration by multiplying with historical runoff, then to flux by multiplying concentration by historical mean annual discharge. Erosion rates were assumed 100 mm ka⁻¹ globally; catchments outside calibration ranges (runoff <150 mm a⁻¹; area >239,000 km²; soil thickness <2.29 m or >22.85 m) were excluded; an area-based extrapolation was applied for 0–20 °C land area (44.5 million km²). Two scenarios were evaluated: SSP1-2.6 (low emissions) and SSP5-8.5 (high emissions).

• Erosion rate as a first-order global control: Normalized alkalinity peaks within an “efficient erosion rate regime” of ~10–1000 mm ka⁻¹, with a maximum near ~100 mm ka⁻¹. Very low (<10 mm ka⁻¹) and very high (>1000 mm ka⁻¹) erosion rates yield low normalized alkalinity, attributed to supply limitation and acid-availability/equilibrium limitation, respectively.
• Areal carbonate proportion: Strong positive control. Highest normalized alkalinity occurs where carbonate is present; carbonate weathering dominates global alkalinity generation. Even non-carbonate-dominated catchments (≤50% carbonate) can yield high alkalinity when other conditions (erosion, MAT) are optimal.
• Temperature (MAT): Temperate climates (≈7.5–15 °C) maximize normalized alkalinity; prior work’s optimum (~11 °C) is supported. Below ~5 °C, low soil respiration reduces acid supply; above ~15 °C, lower carbonate solubility and typical low carbonate coverage reduce alkalinity, with a broader minimum around ~22.5 °C. Model suggests high values in polar/tropical extremes but quantitative interpretation was restricted to 0–20 °C MAT.
• Additional controls: Normalized alkalinity increases with catchment area and soil regolith thickness over the observed range (except very thick regolith >21 m shows low values). Morphology linked to high erosion can enhance turbulence and CO₂ degassing, lowering dissolved alkalinity.
• MAT sensitivity by climate zone: In subarctic/alpine climates (<2.5 °C), glacial/periglacial processes can elevate alkalinity (e.g., more Sr²⁺ from carbonate weathering). Arid/semi-arid warm regions (>15 °C) are water-limited, reducing alkalinity.
• Climate-change impacts (mid-latitudes, historical MAT 0–20 °C): • In the current MAT band 15.0–17.5 °C, alkalinity flux may decrease by up to 68% by 2100, as warming pushes systems toward the modeled minimum (~22.5 °C). Projected average MAT increases within this band: +1.4 °C (SSP1-2.6) and +3.6 °C (SSP5-8.5), yielding mean flux reductions of ~33% and ~68%, respectively. • Temperature bands 12.5–15.0 °C and 17.5–20.0 °C also show decreases in alkalinity flux under warming. • Bands 5.0–10.0 °C show increases in alkalinity flux under both scenarios, driven by enhanced soil respiration; the largest absolute increase is 30.1 t(bicarbonate) a⁻¹ km⁻² for 7.5–10.0 °C under SSP5-8.5. • For 2.5–5.0 °C, flux decreases ~3% (SSP1-2.6) but increases ~11% (SSP5-8.5); for 0.0–2.5 °C, decreases are projected (potentially reflecting deglaciation effects and dataset composition).
• Regional patterns: Greatest increases expected in Central Europe and Central Asia near ~45°N; decreases in parts of North America under SSP5-8.5; reductions north of 60°N and south of 30°N.
• Mid-latitude mean change to 2100: −1.6 t(bicarbonate) a⁻¹ km⁻² (SSP1-2.6) and +0.5 t(bicarbonate) a⁻¹ km⁻² (SSP5-8.5). Using a mean global bicarbonate flux of 19.4 t a⁻¹ km⁻², these imply a global flux change of approximately −8% (SSP1-2.6) and +3% (SSP5-8.5) due solely to mid-latitude concentration changes.
• CO₂ implications (if attributed to carbonate weathering with half the bicarbonate from CO₂): Reduced alkalinity under SSP1-2.6 implies decreased CO₂ sequestration (ΔA ≈ +0.3 tC a⁻¹ km⁻²); increased alkalinity under SSP5-8.5 implies an additional short-term CO₂ sink (ΔA ≈ −0.1 tC a⁻¹ km⁻²). Relative to a global chemical-weathering CO₂ sink of ~−2 tC a⁻¹ km⁻², these represent ~−15% and +5% changes, respectively. In terms of annual anthropogenic emissions, the offset is small (SSP5-8.5 ~0.05%).

The results resolve a long-standing ambiguity by identifying physical erosion rate, along with carbonate presence and temperature, as first-order global controls on riverine alkalinity. Within an efficient erosion regime (~10–1000 mm ka⁻¹), alkalinity generation is maximized, moderated by carbonate availability and temperate MAT that jointly optimize acid supply (via soil respiration), kinetics, and solubility. Outside this regime, either insufficient supply (very low erosion) or acid limitation and enhanced degassing (very high erosion and steep relief) suppress alkalinity. The modeled temperature dependence reconciles prior observations, explaining higher weathering in temperate regions, reductions in cold settings due to low respiration and in warm-arid regions due to water limitation and solubility constraints, and potential glacial/periglacial enhancements where meltwater-rock interactions are intense. Projected warming shifts many mid-latitude catchments away from optimal weathering conditions, decreasing alkalinity flux in currently warm-temperate bands and increasing it in cooler temperate bands, yielding contrasting regional patterns and net changes that depend on emissions pathways. Because alkalinity modulates ocean CO₂ uptake on sub-millennial timescales, these shifts represent a short-term feedback of climate warming on the ocean carbon sink: reduced sequestration under low warming in some regions versus enhanced sequestration under stronger warming in others, contingent on moisture sufficiency. The study underscores the need to integrate geomorphic controls with climatic drivers when projecting future weathering fluxes and carbon-cycle feedbacks.

This work delivers a globally calibrated empirical function linking riverine alkalinity (normalized by runoff) to erosion rate, carbonate coverage, temperature, catchment area, and soil regolith thickness, revealing erosion rate as a first-order control and temperate MAT as optimal for alkalinity generation. It demonstrates that climate warming will substantially alter terrestrial alkalinity fluxes by 2100, with decreases in some mid-latitude bands (especially 15–17.5 °C) and increases in cooler temperate regions, leading to net mid-latitude changes of −1.6 (SSP1-2.6) and +0.5 (SSP5-8.5) t(bicarbonate) a⁻¹ km⁻². These changes imply modest but non-negligible short-term feedbacks on ocean CO₂ sequestration. Future research should: (1) expand datasets to underrepresented regimes (Arctic/high-latitude catchments, very thick regoliths, extremes of runoff and area), (2) better constrain the joint evolution of discharge and concentration under climate change, (3) refine global erosion rate fields beyond uniform assumptions, and (4) assess the roles of morphology-driven degassing and secondary carbonate precipitation in high-relief systems.

• Dataset coverage: Limited by availability of consistent 10Be-derived erosion rates; extremes in predictor space (very low/high values) at mid-latitudes were underrepresented and excluded from flux-change calculations, potentially biasing results.
• Model scope: Quantitative interpretation of MAT effects restricted to 0–20 °C; out-of-range applications (polar/tropical) are uncertain due to unrealistic covariate combinations (e.g., high MAT with high carbonate) and fixed erosion assumptions.
• Assumptions for projections: Erosion rate was uniformly set to 100 mm ka⁻¹; discharge was held constant to isolate temperature effects; catchments with runoff <150 mm a⁻¹ and area/soil thickness outside calibration ranges were excluded (largest 105 basins omitted), with extrapolation applied.
• Hydro-climatic interactions: The combined impacts of changing discharge and weathering on fluxes remain insufficiently understood and were not fully accounted for.
• Regional gaps: Lack of Arctic erosion rate data; model behavior at very high relief may be influenced by turbulence-induced CO₂ degassing and secondary carbonate precipitation not explicitly parameterized.
• Single-model climate input: Warming fields from one ESM (GFDL-ESM4) via ISIMIP were used; multi-model spread and biases were not explored.