Global CO2 emissions from dry inland waters share common drivers across ecosystems

The study addresses the lack of global data on CO2 emissions from desiccated (dry) inland water sediments, which are increasingly prevalent due to natural hydrological variability and anthropogenic pressures (e.g., climate change, water abstraction, reservoir management). Dry inland waters are typically omitted from inland and terrestrial carbon budgets, potentially creating a blind spot in global carbon cycle estimates. The authors hypothesised that CO2 emissions from dry inland waters exceed mean aquatic emissions, making them globally relevant, and that emissions vary with factors influencing CO2 production (organic matter supply, temperature, moisture), gas transport (sediment texture), and geographic properties. They aimed to quantify global CO2 emissions from dry inland waters, identify environmental drivers, and determine whether these emissions resemble aquatic or terrestrial systems to improve global upscaling models.

Previous studies showed exposed sediments can emit CO2 at higher rates than inundated waters, but were limited in scope and geographic coverage. Initial global extrapolations suggested adding desiccated area emissions could increase inland water CO2 emission estimates by 0.4–10%. Many inland waters are subject to temporary flow or seasonal drying, with significant global area experiencing seasonal desiccation, yet these periods are often excluded from inventories and carbon budgets. Prior work has identified small waterbodies as biogeochemical hotspots and highlighted the importance of turbulence in lotic systems for gas exchange. The literature also points to strong influences of moisture and oxygen availability on microbial activity and CO2 production, and potential abiotic contributions via carbonate chemistry. However, globally consistent evidence and quantification of dry sediment emissions and their drivers have been lacking.

Global field survey across 196 dry inland water sites on all continents except Antarctica, spanning rivers, lakes, reservoirs, and ponds, and five Köppen-Geiger climate zones (tropical, arid, temperate, continental, polar). At each site, CO2 fluxes were measured on dry exposed sediments and on adjacent uphill soils (not previously inundated). Three plots per location were measured and averaged per site. Closed opaque chamber method connected to an infrared gas analyser (IRGA) was used; chambers were inserted ~1 cm into sediment or sealed with clay if insertion was not possible. CO2 concentration change in the chamber was tracked for <5 min, and flux (mmol m−2 d−1) calculated via linear regression of d(pCO2)/dt and the ideal gas law: FCO2 = (d(pCO2)/dt) × (R T S) / V, with positive values indicating emission to atmosphere. Sampling avoided vegetated surfaces. Environmental variables (14 total) included in situ or lab-determined measurements: air and sediment temperature; sediment texture (FAO manipulative test); sediment pH and conductivity (1:2.5 sediment:water suspension); gravimetric water content; organic matter content by loss-on-ignition (500 °C) after drying (105 °C). Geographic and climatic descriptors were assigned from global datasets (Köppen-Geiger climate classification; WorldClim annual mean temperature and precipitation). Elevation and latitude were included. Statistical analysis: Linear mixed-effects models (LMMs) using lme4 (lmer) in R 3.4.4 tested the effects of air temperature, organic matter, texture, moisture, conductivity, latitude, elevation, ecosystem type, pH, climate zone, annual mean temperature and precipitation, and second-order interactions among moisture, temperature, and organic matter. The sampling team was included as a random effect (random intercept). Variables were transformed to meet assumptions: log-transform for CO2 flux (x+28), conductivity, organic matter, moisture (x+0.1), elevation; all predictors z-transformed; moisture, elevation, conductivity also log10-transformed as reported. Model simplification removed non-significant predictors. Significance threshold P < 0.05.

- CO2 fluxes from dry sediments ranged from −27 to 2968 mmol m−2 d−1 (mean ± SD = 186 ± 326; median = 93; n = 196). About 4% of sediment sites showed net CO2 uptake; 3% of uphill soils showed uptake. - Emissions from dry sediments were roughly an order of magnitude higher than average lentic water surface emissions (lakes/reservoirs ≈ 27 mmol m−2 d−1) but lower than typical lotic emissions (streams/rivers ≈ 663 mmol m−2 d−1). - Compared to adjacent uphill soils, dry sediment emissions were significantly lower (dry sediments mean 186 vs soils mean 222 ± 277 mmol m−2 d−1; median soils 144; Wilcoxon signed rank test, P < 0.05). Soils had higher organic matter (8 ± 8%) than dry sediments (6 ± 7%). - No significant differences in CO2 fluxes among climate zones; mean ± SD (mmol m−2 d−1): polar 60 ± 58 (median 36), continental 174 ± 140 (125), temperate 178 ± 308 (99), arid 233 ± 470 (61), tropical 236 ± 403 (69). Within temperate sites, dry-winter locations had significantly lower emissions than dry-summer or aseasonal sites (P < 0.05). - By ecosystem type (dry sediments): ponds 267 ± 221 (median 252) > reservoirs 194 ± 478 (82), lakes 215 ± 353 (111), streams 128 ± 218 (64). Ponds were significantly higher than streams and reservoirs (P < 0.05) and marginally higher than lakes. Ponds also had higher organic matter (18 ± 20%) than streams (3 ± 4%, P < 0.05), lakes (14 ± 17%), reservoirs (10 ± 11%). - Variability was higher among sites than among climate zones or ecosystem types, pointing to dominant local controls. - LMM results: fixed effects explained 39% of variance; full model 52%. Strongest predictors (P < 0.001): organic matter content, moisture, temperature, and the interaction organic matter × moisture. Additional significant effects (P < 0.05): interaction temperature × moisture; elevation; latitude; conductivity. Under low moisture, organic matter and temperature had little effect (microbial activity limited by water). High moisture enhanced emissions when organic matter was available; low temperature or low organic matter limited emissions even under high moisture. - Upscaling: global CO2 emissions from dry inland waters estimated at 0.12 ± 0.13 Pg C y−1, equivalent to 6 ± 6% of current inland water CO2 emissions (2.1 Pg C y−1; range 1.56–2.94).

The study demonstrates that CO2 emissions from desiccated inland water sediments are globally prevalent and significant, generally exceeding lentic water surface emissions and approaching lotic emissions but remaining below typical stream/river values. The key drivers of emissions are consistent across climates and ecosystem types, with local moisture, temperature, and organic matter availability, and their interactions, governing microbial respiration and CO2 transport. Including emissions from dry inland waters increases global inland water CO2 emission estimates by about 6%. However, the impact on inventories depends on how desiccated areas and intermittent waters have been treated historically; misclassification can bias estimates up or down. The analysis underscores the need to better map the extent and dynamics of desiccated areas, especially intermittent streams and small ponds that are challenging to detect. The findings suggest emissions may increase in regions facing greater water stress and seasonality under climate change. The study also highlights potentially conservative estimates due to underestimation of desiccated area and exclusion of short-lived but intense rewetting pulses. Understanding temporal variability, vegetation effects, and contributions from abiotic processes is necessary to refine global carbon budgets.

This global survey across 196 dry inland waters shows that CO2 emissions from desiccated sediments are a consistent, significant component of the inland water carbon cycle and share common drivers across ecosystems and climates. Local factors—moisture, temperature, organic matter, and their interactions—govern emission magnitude more than broad climatic or ecosystem classifications. Accounting for these emissions increases global inland water CO2 emission estimates by roughly 6%. Future research should prioritize: improving remote sensing and mapping of temporally and permanently desiccated areas (including intermittent streams and small ponds); quantifying temporal dynamics, especially drying–rewetting pulses; linking emissions to sediment organic matter depletion and vegetation effects; and expanding to other greenhouse gases (CH4, N2O) to capture full biogeochemical impacts of desiccation.

- Geographic sampling imbalance with dominance of temperate sites and underrepresentation of polar regions. - High between-site variability reduces certainty in global upscaling. - Likely underestimation of the global area of desiccated inland waters; intermittent stream and pond areas are difficult to quantify due to canopy/cloud cover and limited detection. - Rewetting events and associated short-term emission pulses were not purposely included and could increase total fluxes. - Potential effects of antecedent conditions (time since desiccation, historical organic matter inputs) and local processes (carbonate chemistry, groundwater CO2 inputs, vegetation and root respiration) were not fully resolved. - Measurements avoided vegetated surfaces, potentially missing contributions from root respiration or plant-mediated fluxes.