The Community Terrestrial Systems Model (CTSM version 5.2) is an advanced land model that simulates physical, chemical, and biological processes in terrestrial ecosystems and climate across varying spatial and temporal scales , in which the Community Land Model (CLM 5) serves as the core land surface component . Land surface heterogeneity is captured through a nested subgrid hierarchy, where each grid cell consists of multiple land units representing lakes, urban areas, crops, glaciers, and vegetated areas. Each land unit includes one or more columns that define the state variables for soil temperature and water content. Columns host patches that represent distinct Plant Functional Types (PFTs) or bare ground for vegetated units, and different crop functional types for cropland units. In total, 16 PFTs are defined, each varying in physiology and structure. In the default CTSM configuration, PFTs may share a soil column, competing for water and energy. In our set-up however, we conducted additional simulations where each PFT resides on its own soil column to investigate the impact of plant competition for water on carbon sequestration scenarios. This setup isolates the effects of transpiration and root water uptake specific to each PFT.

When the crop model is inactive, crops are represented by one irrigated and one unirrigated unmanaged C₃ crop, treated similarly to C₃ grass . Effects of land management on soil carbon, such as harvesting, tillage and erosion are not included in this modeling setup. Land unit distributions are prescribed using the default land cover dataset from the Land Use Harmonization Project 2 LUH2;Appendix Fig. .

The default soil column configuration in CTSM includes 25 layers with a total soil depth of about 50 m. The soil is hydrologically active in the top 20 layers (down to 8.6 m depth). Soil layer depths and thicknesses follow the default CTSM parameterization (Table ). The soil depth to bedrock is spatially variable and prescribed using data from , Appendix Fig. , which constrains the number of hydrologically active layers to the maximum soil depth for each grid cell. One dimensional water flow is described, following the parametrization, by Darcy's law and Richards equation with the hydraulic conductivity (k, mm s⁻¹) and soil matric potential (ψ, mm), which are both varying per soil layer with volumetric soil water (θ, mm³ mm⁻³), soil texture and soil organic matter .

Soil texture and soil organic matter are prescribed based on spatial datasets of clay and sand percentages and organic matter density (kg OM m⁻³). We use the WISE30sec dataset , a harmonized soil profile database based on the World Inventory of Soil property Estimates (WISE), which provides global soil property estimates up to 2 m depth at a 30 arcsec resolution. Percentages of clay and sand from WISE30sec are used as direct input to CTSM (Fig. b, c), while soil organic carbon (SOC, gC kg⁻¹ soil) from WISE30sec is converted to organic matter density (OM, kg OM m⁻³) using the formula: 1OM=SOC⋅ρsoil⋅(100-fcoarse)/100⋅1/0.58 using the bulk density (ρsoil, g cm⁻³) and fraction of coarse fragments (fcoarse, volumetric, %), and 1 g OM is equivalent to the van Bemmelen conversion factor of 0.58 gC (Fig. a). By using SOC and texture values from the same dataset, we ensure consistency across soil characteristics. Within CTSM, organic matter is further converted to an organic fraction using a maximum density of 130 kg m⁻³ corresponding to 100 %, corresponding to the standard organic matter density of peat soils.

The volumetric water content at saturation for soil layer i (θsat,i, mm³ mm⁻³) is computed as a weighted average of the organic and mineral soil components: 2θsat,i=(1-fOM,i)⋅θsat,min,i+fOM,i⋅θsat,OM where fOM,i is the soil organic matter fraction, θsat,OM (mm³ mm⁻³), is the porosity of organic matter set to 0.9; , and θsat,min,i (mm³ mm⁻³) the porosity of the mineral fraction, which depends on the sand fraction (fsand,i) as: 3θsat,min,i=0.489-0.00126⋅fsand, i The volumetric water content at the permanent wilting point (θwp, mm³ mm⁻³) is calculated from the soil water retention function: 4θwp=θsat⋅ψwpψsat-1/B where ψwp=-150000 mm, the soil matric potential at wilting point, B the exponent, and ψsat (mm) the saturated soil matric potential. The exponent B is determined as a weighted average of the organic value (fixed at 2.7) and a mineral value dependent on the clay content (Bmin,i=2.91+0.159⋅fclay,i). The saturated matric potential ψsat is similarly calculated as weighted average, with the organic component fixed at -10.3 mm and the mineral component defined as ψsat,min,i=-10.0⋅10(1.88-0.0131⋅fsand,i). Note that in CTSM, ψsat is therefore not zero but reflects the combined properties of the mineral and organic fractions in each soil layer. Field capacity is defined as the soil water content in each soil layer at a matric potential of -33 kPa. Since the fraction organic matter and the percent sand are prescribed and constant in time throughout a simulation, the volumetric water content at saturation, the field capacity and the wilting point are invariant in time. For a complete description of the parametrization, the reader is referred to the CTSM technical documentation .

In CTSM, soil infiltration is simulated using the Richards equation, where water flow depends on soil hydraulic potential gradients and hydraulic conductivity. SOC affects these processes indirectly by altering soil hydraulic parameters such as porosity, bulk density, and the shape of the water retention curve, which together influence infiltration capacity and vertical water redistribution . Plant transpiration processes, including root water uptake and stomatal conductance, are simulated using the plant hydraulic stress module of . This parametrisation models water transport through vegetation using prognostic vegetation water potential at the root, stem, and leaf levels. Leaf water potential serves as the basis for stomatal conductance water stress, replacing soil potential, while hydraulic root water uptake is simulated using root water potential, replacing the previous transpiration partitioning function . The vertical root distribution used in CTSM follows an exponential decay with depth, resulting in the majority of roots concentrated in the upper soil layers, which are not varying in time (Fig. ).

We conduct land-only simulations for the present day climate using a regular horizontal grid at 0.5° by 0.5° resolution. All simulations are conducted with CTSM version 5.2, using prescribed vegetation phenology based on MODIS satellite observations IHistCLM51SP compset; . The biogeochemistry module is not activated, so soil organic matter remains constant throughout the simulations, and vegetation responses to changes in water availability are not represented. Atmospheric forcing is prescribed by the Global Soil Wetness Project (GSWP3; https://www.isimip.org/documents/405/GSWP3.forcing.HyungjunKim20140306.HJKIM.pdf, last access: 4 June 2026; see also ), a reanalysis product with 0.5° global resolution. GSWP3 provides three-hourly bias-adjusted meteorological variables based on the dynamically downscaled 20th century reanalysis (version 2) of the NCEP model .

The simulation workflow begins with a 60-year spin-up control simulation to ensure steady-state conditions in the soil water compartments. All subsequent simulations branch off from this control and span the period 1985–2014. The first ten years are treated as an additional spin-up, leaving 20 years (1995–2014) for analysis. Throughout the simulations, land use is fixed to the state of the year 2000.

The main land cover types targeted for soil carbon sequestration are cropland. Here, we focus on soil column variables specific to the crop fraction of each grid cell (Fig. ). The crop soil column is not shared by other PFTs. The soil column is not shared by other PFTs. The simulations assume a generic C₃ crop and are conducted without irrigation to isolate the effects of SOC changes on soil moisture and water availability. Since organic soils are unlikely to sequester additional carbon and are not the focus of this study, grid cells with organic carbon contents above 120 gC kg⁻¹ soil are excluded from the analysis.

The effect of carbon sequestration on soil hydrology is assessed by comparing a control simulation with present-day, fixed SOC content to three distinct scenarios representing the soil carbon content after 20 years of active carbon sequestration, rather than a gradual linear increase over time (Fig. d, Table ). A 20-year period reflects a commonly cited saturation point after which a new equilibrium in SOC for the upper soil layers is reached . Thus, the scenarios represent stable states of SOC after 20 years of management. In each scenario, carbon increases are applied to the top 30 cm of the soil column, relative to present-day SOC values. This depth is consistent with prior studies and the 4p1000 initiative, as this layer holds the majority of soil carbon and is influenced by management practices . The input data implicitly assume historical carbon loss through cultivation since the onset of agricultural practices, allowing for soil carbon sequestration to reach a new equilibrium.

The first two scenarios assume uniform soil carbon sequestration applied globally to all cropland grid cells on top of present-day SOC (Table ). These amounts are based on estimates by , who estimated a realistic soil carbon sequestration potential based on the SOC of the SoilGrids 250 m dataset and the scenarios of . The High scenario assumes a total absolute increase of 0.55 %, or 5.5 gC kg⁻¹ soil in SOC over 20 years relative to present-day levels, representing a high sequestration potential for cropland. The Medium scenario applies a total absolute increase of 0.27 % or 2.7 gC kg⁻¹ soil after 20 years. For both scenarios, the increase is applied to all grid cells worldwide. The third scenario, 4p1000, is based on the 4 per 1000 initiative and assumes an annual increase of 0.4 %. Over a period of 20 years, this results in a total increase of 8 % relative to current SOC levels, under the assumption that the increase does not compound annually. So while the first two scenarios represent a spatially invariant, rather small but attainable increases everywhere, the 4p1000 scenario results in spatially variable increases consistent to the pattern of present-day SOC, with high SOC changes in high-latitude regions and lower values in mid to low latitudes (Fig. ). In all scenarios, sequestration is applied to the soil organic carbon (SOC, gC kg⁻¹), based on the organic matter (OM) input layer (Eq. ), for the top five soil layers (0–32 cm), approximating the 30 cm sequestration depth assumption.

The input SOC map with a horizontal resolution of 0.5° by 0.5°, based on the WISE30sec dataset is validated by comparing the mean SOC in the top 30 cm of cropland soil with values from literature. The mean SOC in cropland topsoils amounts to 68 t ha⁻¹ and is derived using the SOC (gC kg⁻¹) and soil bulk density (g cm⁻³) of the CTSM input derived from WISE30sec and a soil depth of 30 cm. This value aligns with recent studies: reports 1.3 gC kg soil⁻¹, corresponding to 59.1 t ha⁻¹ derived using a neural network based on global climate, land cover and soil datasets. cites 82 t ha⁻¹, based on the Soils Grid 250 m global database. Additionally, estimates total cropland SOC at 131.81 PgC, while our CTSM input map yields 108.42 PgC using the same cropland mask.

Four experiments are conducted, which differ in the soil organic matter input provided. The control scenario (CTL) uses the default WISE30sec SOC map, while the High, Medium, and 4p1000 scenarios use modified soil organic matter input maps.

The effects of carbon sequestration on soil hydrology are quantified by comparing differences between the scenarios and CTL simulations across various soil moisture-related variables including water holding capacity, saturated fraction, volumetric and total water content. To this end, we compare mean values from the 20-year simulation period across experiments (Sect. ). The water holding capacity is defined as the plant available water, calculated as the difference between field capacity (θfc, mm³ mm⁻³) and wilting point (θwp, mm³ mm⁻³).

The saturated fraction (θs, mm³ mm⁻³) represents the extent to which the soil is saturated and is calculated as: 5θs=θθsat where θ (mm³ mm⁻³) is the actual volumetric water content, and θsat (mm³ mm⁻³) is the volumetric water content at saturation.

Soil water stress conditions are defined based on a threshold relative to field capacity and the wilting point. Water stress occurs when the volumetric soil water content (θ) falls below 50 % of the water content at field capacity (θfc), provided that this threshold remains above the wilting point (θwp). In cases where 0.5⋅θfc<θwp, the wilting point is used as the stress threshold instead. The annual water stress is then quantified as the cumulative deficit between the applicable stress threshold and the actual soil water content (θ), accumulated over the year and summed across the first seven soil layers (d=0.68 m), representing the upper 60 cm of the soil profile corresponding to the depth affected by irrigation (Eq. ). 6θstress,i=max⁡0.5⋅θfc,i,θwp,i7water stress=∑i=17∑month=112(θstress,i-θmonth,i)⋅dforθmonth,i<θstress,i

The impact of soil carbon sequestration is measured through changes in water holding capacity, saturated fraction, and volumetric water content of the High scenario compared to CTL (Fig. ). Water holding capacity in each simulation remains constant over time, as defined by the model formulation. Spatially, however, it increases globally, with a global mean volumetric rise of 0.002 m³ m⁻³, or 2 mm water per 100 mm soil (Fig. a). This pattern is driven by a consistent increase in soil water content at field capacity, strongly influenced by the soil organic carbon (SOC) fraction (Appendix Fig. b) and reflects the model's representation of the improved soil porosity from added organic carbon. While soil water content at wilting point also increases, this effect is less widespread and less pronounced (Appendix Fig. a). Changes in the saturated fraction exhibit a more heterogeneous spatial pattern and are generally of much lower magnitude, with areas showing substantial decreases (Fig. b). These decreases occur when soil water content at saturation (θsat) increases more than the mean simulated volumetric soil water content (θ), averaged over all years. This mean volumetric water content generally increases when weighted over the soil layers, with a global mean increase of 0.002 m³ m⁻³ (Fig. c). Overall, the changes are small and consistent in the High scenario, which involves a small forcing of a 5.5 gC kg⁻¹ soil (0.55 %) increase in SOC.

With the soil's increased capacity to retain water and rising volumetric water content in most regions, a key question emerges: Is this additional water accessible to plants, which would reduce water stress and thereby potentially alleviate pressures on irrigation water abstraction and unsustainable use? Given the importance of this issue, and the potential of soil carbon sequestration for croplands, the remainder of our analysis focuses on the crop fraction of the grid cell. Cropland resides on its own soil column, which elucidates the competing effects of different PFTs with different rooting depths.

The effect on mean total water availability and total soil liquid water content shows an increase of 2 mm averaged globally, and varies by region (Fig. a); hereafter, soil water content refers to the liquid fraction. In areas such as the Amazon, the eastern United States and Canada, Scandinavia, Western Europe, West and Central Africa, Madagascar, and East and Southeast Asia, soil carbon sequestration leads to an increase in total soil water content. Conversely, a marked decrease in soil water content is observed in regions such as the western United States, the Argentinian Pampas, parts of Southern Africa (spanning Namibia, Angola, Botswana, and South Africa), the Sahel, areas of East Africa, portions of the Eurasian continent, northeastern China, and Australia. These contrasting responses result from differences in soil water content above and below a depth of 32 cm.

In the upper 32 cm of the soil, where soil carbon sequestration is applied, mean water content consistently increases across all crop regions, with a global average increase of +4 mm (Fig. b), likely reflecting a redistribution of water that is retained in the topsoil and therefore less available for percolation to deeper layers. In contrast, the deeper soil layers below 32 cm exhibit diverging patterns (Fig. c, global average decrease of -1 mm). In most cropland areas, water content in these deeper layers decreases, particularly where total water content is reduced throughout the entire soil column. However, some regions – such as eastern Canada, Scandinavia, and parts of India and Southeast Asia – show increases in deeper layer water content. These increases may be linked to ice melting within the model or regional soil characteristics. The soil depth in these layers is determined by the soil depth map (Appendix Fig. ), which defines maximum depths for each location. Areas with strong declines in water content tend to correspond with more arid or sandy regions (Fig. b), indicating that water is retained in the upper soil layers rather than percolating into deeper depths.

To better assess regional differences, we examine volumetric water content (θ) and the saturated fraction (θs) in three IPCC AR6 regions: Western North America (WNA), Northeastern Africa (NEAF), and South Asia (SAS) Fig. a; . WNA, a dry region with extensive irrigation and a rainy season from November to March, has 50 % sand and 20 % clay (Fig. b, c). NEAF, covering the arid Horn of Africa and wetter Ethiopian Highlands, has high sand (55 %–60 %) and low clay (<20 %) fractions, with a dry season from December to March. Both regions show no clear increase in total water content due to declines below 32 cm (Fig. a, c). SAS, a major irrigated region with a monsoon season from April to September, has 30 % sand and 60 % clay, leading to a slight increase in soil water content below 32 cm (Fig. c).

Seasonal variations in the influence of soil carbon sequestration on soil water content for these regions are examined through vertical profiles of changes in volumetric water content and the saturated fraction (Fig. ). Volumetric water content is highest between 20 and 32 cm – the deepest layer where soil carbon sequestration is applied (Fig. a–c). Across all three regions, water content increases most during the rainy season and persists at greater depths beyond the wet periods. In contrast, a slight soil carbon sequestration-induced drying occurs in the surface layers during the dry season, with the deepest drying observed in Western North America. Below the carbon-sequestered layers, changes in water content are minimal, with slight drying or no change observed. This pattern suggests that increased SOC in the upper layers enhances water retention in the upper layers, thereby reducing percolation and limiting the downward movement of water through soil texture effects, increasing water retention.

The reduction in saturated fraction is most pronounced in the upper layers and diminishes with depth, indicating lower surface soil saturation (Fig. ). Relative increases in volumetric water content reach up to 10 % at the deepest soil carbon sequestration layer, with minimal seasonal variation compared to absolute values (Appendix Fig. ). This indicates that soil carbon sequestration enhances water retention within the sequestration layers, thereby amplifying soil water seasonality. At the same time, it reduces the downward percolation of water into layers below the soil carbon sequestration depth. Especially in the surface layers, the increase in saturated fraction is not fully matched by the actual increase in water content from percolation, resulting in less saturated surface soil layers.

To assess whether the increased water content in upper soil layers improves plant water availability, we examine its impact on evapotranspiration components: soil evaporation, vegetation transpiration, vegetation evaporation, and the total evapotranspiration (Fig. ). Carbon sequestration has a clear spatial effect on soil evaporation, but its impact is smaller compared to vegetation transpiration (Fig. a). Soil evaporation increases in more sandy regions like the Monte Desert in Argentina, Western Southern Africa, Western North America, as well as the Sahel and parts of the Eurasian Steppe. In contrast, areas with higher clay fractions, such as the Ethiopian Highlands, parts of India and Northwestern Australia, show reductions in soil evaporation. Soil carbon sequestration seems to reduce soil evaporation in clay-rich regions while enhancing it in sandy ones. The increase in soil evaporation appears to reduce the water available for transpiration. This is somewhat unexpected, given that evaporation primarily draws from near-surface moisture, whereas transpiration accesses water from deeper soil layers.

Soil carbon sequestration generally boosts vegetation transpiration, suggesting that plants have more access to water (Fig. b). This effect is particularly notable in clay-rich regions, such as in India, Southeast Asia and the tropical rain forests. In these areas water is typically not limiting, and the effect likely reflects local increases in soil water retention and root-zone moisture storage rather than a true alleviation of water stress. In contrast, the impact on vegetation evaporation, i.e., evaporation from precipitation intercepted by the canopy, is minimal (Fig. 5c). This can be explained by the fact that this flux is independent of soil water content. However, because 2 m temperature is a diagnostic variable in CTSM, a small temperature-related feedback is observed in vegetation evaporation rates. The overall change in evapotranspiration is primarily driven by increased vegetation transpiration, with a smaller contribution from enhanced soil evaporation and amounts up to a global average increase of +2 mm (Fig. d).

Annual water stress decreases across most regions, particularly in Western North America, parts of South America, the Sahel, Western Southern Africa, South Asia and the Middle East, which correspond to areas with high sand contents. The decrease in water stress reaches several meters per year in some areas (Fig. 6a), indicating that carbon sequestration reduces the accumulated soil moisture deficit below the stress threshold over the annual cycle, particularly in coarse-textured regions. These results shows the potential for reducing irrigation demands. It is however not possible to quantify the direct reduction in irrigation needs due to the way irrigation is parametrized in the CTSM, using a threshold soil moisture value that also changes with increased soil carbon.

Soil carbon sequestration reduces surface runoff in most regions (Fig. a, -1 mm globally averaged), with pronounced decreases observed in Scandinavia, Central Europe, northeastern Canada and the western Amazon. This reduction suggests improved infiltration rates (Appendix Fig. ). Subsurface drainage, which refers to water exiting the bottom of the grid cell, also decreases in most areas, except in regions like northeastern Canada, Scandinavia and Central Europe (Fig. b, -0.6 mm globally averaged). The simulated reduction in drainage is attributed to a lower saturated fraction across the soil column (Sect. , Fig. ), leading to reduced water tables and subsurface runoff. This result aligns with the model’s representation of subsurface drainage . Overall, these findings suggest that increased soil carbon in CTSM generally enhances the soil's ability to retain water.

The soil water balance responses described above correspond to the High soil carbon sequestration scenario. The Medium soil carbon sequestration scenario exhibits similar spatial patterns but with smaller magnitude differences (Appendix Figs. , , , and ). The 4p1000 scenario, where soil carbon sequestration is relative to existing carbon stocks (Appendix Fig. ), shows distinct spatial patterns, which are reflected in the hydrological responses (Appendix Figs. , , , and ).

Comparing volumetric water content across the three selected regions (as shown in Fig. a) shows compensatory effects, with increased soil moisture above 32 cm and decreases below (Fig. ). This pattern is most pronounced in Western North America and Northeastern Africa, where both the High and Medium soil carbon sequestration scenarios result in a net soil water increase. However, in the 4p1000 scenario, the decline below 32 cm outweighs the gains above, leading to an overall reduction in volumetric water content. In South Asia, decreases below 32 cm are minimal, and in the 4p1000 scenario, a slight increase is observed, resulting in a strong overall increase in soil water in the full column (Fig. ).