We conducted a literature review to identify studies that examined changes in soil properties resulting from management adaptations aimed at increasing SOC (Table ). Estimates of soil hydraulic properties – such as soil moisture at field capacity (θFC), permanent wilting point (θPWP), or saturated hydraulic conductivity (Ksat), derived from pedotransfer functions (PTFs) can vary considerably and are a source of uncertainty . However, since PTFs are (ideally) trained on large soil datasets from similar pedoclimatic conditions, they should support a broad generalization and enable the prediction of difficult-to-measure parameters from more easily observable ones. Moreover, they typically cover a wider range of soil textures than field or experimental studies.

A range of management practices have been shown to increase SOC: including cover cropping, diversified crop rotations, the application of organic amendments (e.g., compost or manure), the retention of crop residues and the application of biochar (Table ). These practices are often combined, and the magnitude of SOC increase varies depending on site-specific conditions, depth, and implementation duration. The reported increases in SOC range from 7 % to 36 %, 20 % to 220 %, and 60 % to absolute increases of approximately +1 % by mass .

In addition to changes in SOC, several studies report concurrent effects on other soil hydraulic properties. A reduction in bulk density (ρb) is frequently observed; the effect varies from -1 % to -4 % through cover cropping to -14 % for a silt loam in response to long-term organic amendments .

The ranges of change in saturated hydraulic conductivity (Ksat) are particularly variable, with reported increases of 50 % to 250 %, 40 % to 360 % and 95 % depending on the practice and the site. Ksat underlies large variability and is a generally hard-to-measure and even harder-to-estimate variable and should be handled with caution .

Soil moisture (θ) and especially plant available water capacity (PAWC), are reported to increase in the range of 4 % to 20 %, 4 % to 54 %, 33 % and 65 %, following increasing SOC and decreasing ρb .

In , reducing crop residue cover from 100 % to 0 % decreased SOC, increased ρb, and reduced Ksat and PAWC. In Table , we assume that increasing residue cover from 0 % to 100 % would have the opposite effects: increasing SOC, reducing BD, and increasing Ksat and PAWC.

The role of tillage is more complex. While reduced or no-tillage is often associated with higher SOC in the topsoil, it primarily leads to a redistribution of organic matter, with less SOC in deeper layers, and total SOC differences are not always significant . Tillage is often used in organic farming to control weeds, which can offset some of the beneficial effects of organic farming practices. In particular, describe how tillage can negatively affect bulk density (ρb) and Ksat, potentially counteracting the positive impacts of increased SOC in organic management systems, depending on tillage frequency and intensity.

The mesoscale Hydrologic Model (mHM, v. 5.13.1; https://mhm-ufz.org, last access: 10 May 2026) is a spatially distributed, process-based model designed to simulate major hydrological processes and water balance across diverse hydroclimatic regions and scales . The computation of soil moisture processes and the generation of mobile water takes place at a grid scale, followed by a HBV-like soil moisture-runoff transformation to transform grid-scale mobile water to grid-scale runoff, followed by transfer and routing from grid cell to grid cell following topography-based flow directions (see below). The multiscale-parameter regionalization (MPR) is a key feature of mHM, which allows for both high-resolution spatial input data and computational efficiency . Using transfer functions, effective model parameters (such as hydraulic conductivity) at the grid-scale are estimated from spatial input parameters such as soil texture. These effective parameters are then internally upscaled to the (coarser) model resolution using different operators such as harmonic or arithmetic mean, while retaining spatial variability . More detailed descriptions are available in the work of with more specific details on soil hydraulic parameterizations in .

The main processes simulated in mHM are canopy interception, snow accumulation and melt, evapotranspiration, infiltration, soil moisture storage, surface runoff, lateral subsurface flow (called interflow in mHM), percolation, groundwater storage, baseflow and in-stream routing (Fig. ). Snow accumulation is simulated with a simple temperature threshold; snowmelt is based on a degree-day method. In mHM, surface runoff can only occur on (nearly) impervious grid cells representing sealed areas such as streets or buildings. Potential runoff from excess water is assumed to re-infiltrate at the grid-scale and is, therefore, not simulated as a separate process in mHM. This is justified by the typically recommended grid resolution of 1 km to 50 km .

The soil moisture and runoff generation schemes in mHM are conceptually based on the HBV model , with some differences: mHM simulates soil moisture dynamics per soil layer (HBV usually has a single layer); the routine is described in more detail in Sect. . Once mobile water is generated per soil layer, the HBV conceptualization is used to transform grid-scale mobile water into grid-scale runoff. Each grid cell uses two subsurface storages fed with the sum of the mobile water from the soil moisture routine. The upper unsaturated storage generates faster responding interflow and the other slower responding baseflow (Fig. ). Fast interflow occurs only if the water level in the storage zone exceeds a threshold; slow interflow is a permanent flux governed by the water level in the first bucket . The remaining water level in this zone is the base for the percolation flux, encoded as a linear function of the water level. The percolation feeds the deeper saturated zone, supposed to emulate groundwater storage, where baseflow is again parameterized as a linear function of the water level .

The total generated runoff (interflows and baseflow) from each grid cell is routed through the modelling domain by the multiscale Routing Model (mRM), a key component of the model . Grid-scale runoff is transferred from cell to cell following topography-based flow direction and flow accumulation map. The routing algorithm applies the kinematic wave equation with spatially varying flow celerity parameterized by slope . An adaptive time-stepping scheme is used to ensure numerical stability across resolutions. developed the subgrid catchment conservation (SCC) routine specifically for mHM as an alternative to the commonly used D8 algorithm . This approach addresses the catchment size problem that arises when small catchments are simulated at coarse resolution, which can lead to over- or underestimation of catchment area and the resulting streamflow. For cells intersecting several subcatchments, SCC allows water to partition into different neighboring cells. Due to this study's relatively small catchment size, we also employ the SCC algorithm, which reduces biases in discharge between different subcatchments .

In the configuration chosen for this study (see Sect. ), the model has 47 (global) parameters that are calibrated based on observed streamflow (calibrated parameter values are shown in Supplement Sect. S4, Table S3). mHM has a built-in calibration algorithm based on a dynamically dimensioned search algorithm for single objective parameter optimization. The users can choose between several performance criteria (https://mhm-ufz.org, last access: 10 May 2026). The retained calibration options for the case studies at hand are further discussed in Sect. . The number of iterations is set to 2500, which has been successfully used to calibrate the mHM model in other studies .

The mHM model represents root-zone soil moisture dynamics across multiple soil layers, with each layer corresponding to an individual soil water reservoir. The water balance within each reservoir is primarily controlled by incoming fluxes – snowmelt and rainfall in the uppermost layer, or percolation from the overlying soil layer in lower layers – and outgoing fluxes, including downward percolation and layer-specific evapotranspiration. Each soil layer has an upper soil water limit, represented by θsat, which acts as a threshold for storage capacity. θsat is estimated using the PTF by (mHM default): 1θsat=Cconstant+Cclay⋅τclay+CDB⋅ρb where Cclay is the clay content, and Cconstant, τclay, and CDB are (global) parameters that are calibrated (Sect. S4, Table S3). At each time step, the current water content θ in each soil layer is compared to θsat; if θ is below saturation, infiltration into the layer is allowed. A portion of the incoming water is retained in the current layer, while the remainder percolates into the next layer (see Equations in Appendix ). This also means that if θsat (i.e., the soil's water retention capacity) increases, then for the same water input, less water infiltrates to deeper layers.

In the default mHM setup, bulk density (ρb) is internally estimated from a user-defined mineral bulk density and modified using an organic matter parameter (pOM), which can be fixed or calibrated but is spatially uniform (Fig. ). Saturated hydraulic conductivity (Ksat) in mHM is derived using the pedotransfer function (PTF) from , based on sand and clay content. We modify this parameterization to evaluate the effect of different SOC scenarios by directly linking SOC to ρb and Ksat (Fig. ). Specifically, we bypass the internal pOM routine and instead input SOC-adjusted ρb values directly, using the PTF from , adapted by : 2ρb=1.660-0.318τSOC, where τSOC is the SOC content. Here we follow who showed that SOC consistently affects ρb in a largely texture-independent way. This PTF was trained on the extensive USDA soil database and is therefore assumed to be transferable to our study region. By representing SOC changes through ρb in pedotransfer functions, the resulting soil hydraulic parameters naturally reflect SOC effects .

Above mentioned adaptation also allows us to capture the observed relationship between increasing SOC and decreasing ρb, which is generally linked to higher Ksat . To incorporate the effect of increasing SOC onto Ksat, we also replace the default PTF with the one proposed by , as listed in , which includes SOC and ρb as predictors: 3Ksat=CKsat1exp⁡(CKsat2-CKsat3ln⁡(τclay)-CKsat4ln⁡(τsand)-CKsat5ln⁡(τSOC)-CKsat6ρb), where τclay and τsand are the clay and the sand content, the parameters CKsat1 to CKsat6 are constants (values listed in Sect. S4, Table S2). This PTF was trained on a Belgian database which includes soils present in our study region. In mHM, θFC is parameterized as a function of Ksat, such that higher Ksat corresponds to lower θFC, based on the PTF derived from soil database analysis by .

These parameter adjustments propagate through the process simulation chain, and their effects on parameters, variables, states, and fluxes in response to increased SOC will be described and illustrated in the results Section (Fig. ). They influence the estimation of the van Genuchten parameters used to compute θsat, α, n, and m, as well as field capacity (θFC) and the permanent wilting point (θPWP) (Equations listed in Appendix ). These, in turn, affect the simulated soil moisture (θ) and the associated fluxes, including infiltration, evapotranspiration (ET), lateral subsurface flow, and percolation.

ET in mHM is computed as a reduction of potential evapotranspiration (PET) by a soil moisture stress factor, following the formulation of or . In this study, we used the mHM process representation of , which combines the Jarvis approach with a root distribution model based on . In this configuration, root density varies spatially and vertically as a function of soil field capacity (θFC).

The reduction from PET, after accounting for canopy interception, to ET is expressed as: 4ET=PET⋅f where f is a soil moisture stress function defined by: 5f=R,θ‾≥tjarvisRθ‾tjarvis,θ‾<tjarvis

Here, tjarvis is a calibrated threshold parameter, θ‾ is the normalized soil water content: 6θ‾=θ-θpwpθsat-θpwp, and R is the fraction of roots in each soil layer: 7R=1-RCoeffFCdu-1-RCoeffFCdl with RCoeffFC representing the root fraction coefficient for the layer, and du and dl denoting the upper and lower soil layer boundaries (Appendix ). This formulation allows soil-layer specific root fractions to modulate ET in response to soil moisture.

We apply the mHM model to the mid-sized (602 km²), lowland, pluvial Broye catchment in Western Switzerland (Fig. ). The modeling period is constrained by the availability of leaf area index input data and is therefore set to 2015–2022, with 2015 used as a warm-up period and discarded from the analysis. Despite the relatively short study period, there is considerable variability, with 2018 and 2022 as hot and dry years, 2017 as dry, 2016 and 2021 as cool and wet years, and some intermediate years (2019 and 2020, Fig. ).

The mHM model is set up for the entire Broye catchment domain, but discharge observations are available only for four nested subcatchments. One of these subcatchments is also named Broye and drains the largest area, with its outlet near the city of Payerne (Fig. ). For clarity, we refer to the full modeled domain as the Broye catchment and to the gauged subcatchments as Broye (subcatchment), Flon, Arbogne, and Petit Glâne.

The landscape is dominated by cropland interspersed with small forest patches. Soils are primarily loams, clay loams or sandy loams, with rather low SOC contents (averaging at 2.2 % in the topsoil, as shown in Fig. ). The region has a temperate climate, with mean annual precipitation of 1142 mm and mean annual temperature of 9.12 °C (1993–2022). The streams exhibit a typical pluvial flow regime, with discharge peaks in winter and low flows in summer, characteristic of lowland Swiss agricultural catchments.

The required input data and their sources are listed in Table . The morphological and land use input data have a resolution of 50 m × 50 m, and the meteorological data of 1 km × 1 km, which is also the internal modeling resolution. The water transfer and routing in the model are based on the provided flow direction. However, because the water flow in the flat part of the catchment is not well constrained by the DEM, a reconditioned DEM consistent with the mapped rivers must be calculated. After trying different tools that provided unsatisfactory results, we developed a new tool to seamlessly align DEMs with mapped stream networks, resulting in minimal terrain alteration: hydro-snap . The approach is softer than a stream burn-in and alters the DEM only where necessary. It also constrains the flow direction to be consistent with a provided catchment boundary.

With the available gridded precipitation data , the water balance in the subcatchments Petit Glâne and Arbogne does not close (Appendix ). The observed annual discharge is far too low compared to the catchment-average precipitation. However, comparable catchments nearby show similarly low discharge values ; accordingly, discharge measurement errors alone cannot explain the difference. The gridded precipitation product we use might well contain interpolation artifacts given the substantial spatial variability of observed precipitation. Therefore, we also explored other precipitation products (Sect. S3). To reduce potential biases, we eventually combined the gridded precipitation product with data of the nearby meteo stations for the two smaller subcatchments, Arbogne and Petit Glâne (Sect. S3, Fig. S5).

Different options are available to represent the hydrological processes in mHM (see https://mhm-ufz.org, last access: 10 May 2026, for details). We select the default options , except for the soil moisture and the evapotranspiration routine. For the soil moisture routine, we select the option where ET in each soil layer is regulated by the relative available soil moisture, rather than being uniform across land use classes, implemented by . This option allows for a spatially varying root fraction distribution depending on the soil's field capacity (θFC), which is an advantage in the presence of a high-quality soil database of high resolution (90 m × 90 m) . In contrast to most crop and land surface models, where root distribution is prescribed as a depth-dependent function independent of soil moisture , mHM explicitly links the root distribution to the soil's θFC following .

We compute PET as an external model input according to the Priestley-Taylor method , which uses average temperature, solar radiation and elevation as input. We set the model options such that PET is further distributed in space based on aspect, as implemented by . It should be noted that the mHM option to correct PET based on LAI data led to unrealistically high PET/ET values in our case and was thus not used.

The model is calibrated using as performance criterion the Kling-Gupta efficiency (KGE) , calculated on each of the observed streamflow time series (at the four gauges) and averaged thereafter (without weighting). The retained period for calibration is 2016–2019 and for evaluation 2020–2022. The specific calibration setting is the result of manual explorations of different objective functions and of number of iterations. With fitting the model based on the KGE, we could get the overall best performance while maintaining realistic dynamics of all states and fluxes. We also evaluate the model performance for soil moisture using observed timeseries of volumetric water content at three depths from a grassland site close to the gauging station of the Broye subcatchment in Payerne, measured as part of the Swiss Soil Moisture EXperiment SwissSMEX . The model is run at a daily timestep.

We apply different scenarios to evaluate the effect of increased SOC on catchment hydrology to (i) represent possible outcomes from long-term agricultural management adaptations (Sect. ), and (ii) test the model's sensitivity towards different levels and depths of SOC increases (Fig. ). In the Broye catchment, SOC is on average around 2.2 % in the first 30 cm (soil layer 1), and approximately 0.9 % between 30 and 60 cm (layer 2, Fig. ). The SOC ratio between layer 2 and layer 1 is therefore approximately 60 %. We apply this depth-decrease ratio to the increase scenarios 1, 3, and 5. These scenarios represent increasing magnitudes of SOC increases. In scenario 2, SOC is not increased in soil layer 2 at all, and in scenario 4, SOC is increased by 1 % (mass) in both layers.

We emphasize that these scenarios are artificial and not intended to represent specific, immediately achievable management interventions, but should rather reflect the long-term possible outcomes of combinations of different management adaptations. While they are informed by the literature review (Sect. ), the scenarios with large and deep SOC increases (MedC_MedC and vHighC_MedC) may be harder to achieve in practice. Nevertheless, including such scenarios allows us to explore the potential range of hydrological responses to SOC increases and test the sensitivity of the model to large changes in soil properties.

The SOC values in each scenario are then used to estimate the model input ρb (bulk density), as discussed in Sect. . Consistent with the rationale outlined in the introduction, we hypothesize that increasing SOC will enhance soil water retention, allowing the soil to buffer hydrologic extremes, reduce low-flow frequency, and modestly attenuate peak flows. mHM considers three land use types: forest, impervious cover and pervious cover, where the latter includes all cropland and meadows. The adaptations to ρb are only applied to pervious areas, which have the highest share in each subcatchment (shown in Fig. , panel a). Within the pervious fraction, the actual land use composition differs across catchments: Petit Glâne and Arbogne are cropland-dominated (80 % and 66 % of pervious area), the Broye subcatchment is mixed (54 % cropland, 43 % permanent meadow), and the Flon is dominated by permanent meadows and pastures (80 % of pervious area).

To assess the impact of the SOC scenarios on low flows, we calculate the Q347 threshold for each subcatchment. Q347 corresponds to the discharge that is exceeded on 347 d yr⁻¹ (i.e., the 5th percentile) and is commonly used in Switzerland to define low-flow periods and as a threshold for the restriction of irrigation water withdrawal from rivers .

For peak flows, our analysis is constrained by the daily resolution of simulated discharge, whereas hourly peaks would be more relevant . Nevertheless, we estimate changes in discharge associated with two-year return period floods (Q2 events). Q2 thresholds are determined for each subcatchment using a generalized extreme value model, although the short time series in smaller subcatchments can be limiting (48 years in the Broye subcatchment and 26 years in the other subcatchments). Across stations, we also observe a decreasing discharge trend, significant only for the Broye subcatchment, which explains why only a few Q2 events occur during the study period.

Model calibration led to a good fit of simulated to observed streamflow for the Broye subcatchment and the Flon (KGE = 0.91), with a slightly lower performance for the Arbogne and Petit Glâne (KGE = 0.83 and 0.86, Fig. ).

Seasonal discharge dynamics are not equally well captured across subcatchments (Appendix  and ). The Broye subcatchment shows the best fit; low flows are underestimated in the Flon, overestimated and mis-timed in the Arbogne, and mis-timed in the Petit Glâne. Percentage biases for high and low flows (Q95 and Q5) range from -0.1 % to 7 % and -22 % to 43 %, respectively, with the best agreement in the Broye subcatchment (pbias for Q95 = 7 %, pbias for Q5 = 9 %), likely reflecting the higher quality of observed discharge data there.

In comparison with the observed soil moisture time series (Sect. ), mHM achieved reasonably good performance, except for the lowest soil layer, with KGE values of 0.65, 0.73, and 0.13 (0–30, 30–60 and 60–90 cm). The good fit in the two upper layers is noteworthy given that the data were not used for calibration and represent a single grid cell. While soil moisture was generally underestimated (percentage bias 8 % to -11 %), the model reproduced the temporal variability well (Appendix ).

In Fig. a, we show the impact of the SOC increase on several parameters of interest. The points represent all pervious land cover cells in the overall catchment, which equals the area the SOC increase is applied to. Saturated hydraulic conductivity (Ksat) and water content at saturation and wilting point (θSat and θPWP) are effective model parameters calculated within mHM. θFC changes with the almost same magnitude as θSat, which is why we do not show it explicitly in Fig. . Bulk density (ρb) is a model input and PAWC (plant available water capacity, θFC-θPWP) gives an idea if the surplus in retained water could be taken up by plants.

Adding +0.5 %, 1 % and 1.5 % SOC to the first soil layer led to a decrease in ρb by on average 3.3 %, 6.4 % and 9.2 % (Fig. ). The decrease in ρb propagates through the model (Fig. ), leading to averages increase by 3.2 %, 6.2 % and 8.8 % in both θSat and θFC. PAWC is increased by 4.9 %, 9.3 % and up to 13.5 %. As described in Sect. , an increase in Ksat leads to a decrease in θFC, as parameterized in mHM. Although Ksat was substantially increased, the effect on θFC is negligible. The sensitivity of PAWC to increases in SOC depends on the initial SOC content and soil texture. PAWC increases more strongly in soils with low initial SOC and higher sand content (Sect. S7).

We first isolate the grid cells where SOC-induced changes in soil hydraulic properties were applied (= pervious landcover cells) to examine how these properties change and how states and fluxes respond locally. Subsequently, we aggregate the results to evaluate how these local changes propagate to influence hydrological states and fluxes at the catchment scale.

Figure b illustrates how changes in hydraulic parameters propagate through the process chain in mHM and how states and fluxes are changed accordingly, here representing a snapshot for the simulated net changes. On average, increased soil water retention capacity leads to slightly higher ET and since more water can be retained and evaporated, less water contributes to further states and fluxes downwards.

The overall impact of the SOC scenarios at the grid-scale are moderate. Figure  shows actual evapotranspiration (ET), soil water content in the first and second soil layer, and subsurface runoff across all pervious land use cells in the overall catchment (other states and fluxes are shown in the Sect. S2). Subsurface runoff in mHM comprises the fluxes fast and slow interflow and baseflow. Figure  displays the range over all cells and the mean in solid lines: panel (a0 shows the timeseries, (b) shows the relative differences to the base scenario and panel (c) shows the cumulative differences. Average soil water content in layers 1 and 2 increases by average 2.9 % to 8.1 % (over all SOC scenarios), corresponding to 3–8 mm in winter and 2–6 mm in summer, with substantial spatial variability (Fig. ). The impact of SOC increase on the boundary fluxes ET and subsurface is smaller: ET increases slightly (+0.16 % to +0.4 %) while subsurface runoff slightly decreases (−0.28 % to −0.72 %), corresponding to 0.2–0.6 and 1–2 mm, respectively. The differences in these key state and fluxes exhibit distinct seasonal patterns. For ET, differences peak in spring and summer, subsurface runoff peaks in winter are partly reverses in summer and fall. The difference in soil water content is largest in winter and spring, decreasing in late summer and early fall, before it sharply rising again in late fall.

We summarize the average seasonal pattern of all SOC scenarios relative to the base scenario in Fig.  and can distinguish four stages, defined in Table , which outline the main hydrological responses throughout the year. Overall, we see a consistent increase in soil moisture across all seasons, moderate increases in ET during spring and summer, and generally reduced subsurface runoff, except in spring when it shows a slight increase (Table ).

Table  shows the relative changes for ET, total soil water content, and subsurface runoff across all cells in each subcatchment and the entire catchment over all scenarios. While overall changes are small, the signal of increased SOC is clearly visible at the catchment scale, largely due to the dominance of agricultural land. Among subcatchments, Petit Glâne and Arbogne show the largest relative changes for all variables, whereas Flon and Broye subcatchment exhibit smaller responses. Subsurface runoff at the catchment scale corresponds to the river discharge; thus, the changes in this variable reflect the impact of SOC increases on overall catchment discharge.

Beyond grid-scale subsurface, mHM also simulates routed discharge at the locations of gauging stations. The overall effect of increased SOC on discharge is small. Because hydrographs from all scenarios almost entirely overlap, they are shown only in the Sect. S6. Relative differences between the base and SOC scenarios are moderate (positive values indicate higher discharge under SOC; Fig. ), which is consistent with the small changes in subsurface runoff at the grid-cell scale (Sect. ).

All subcatchments display a similar seasonal pattern: the relative difference in discharge decreases mostly in fall and winter, and increases in spring and summer. However, the magnitude and direction of changes differ by subcatchment (Figs. b and ). Across catchments, relative discharge responses vary in both magnitude and direction. The Flon shows the strongest increases, with values reaching up to +20 %, while the Petit Glâne exhibits the largest decreases of up to -18 %. For most catchments, relative differences remain within ±10 %. In absolute terms, the Arbogne (mean discharge (mean Q) 0.73 m3s-1) shows increases of up to 0.08 m3s-1 and decreases between 0.10 and 0.30 m3s-1. In the Broye subcatchment (mean Q 7.89 m3s-1), discharge can increase by up to 0.8 m3s-1 and decrease by 1-3.5 m3s-1. For the Flon (mean Q 0.34 m3s-1), increases reach up to 0.05 m3s-1, while decreases range from 0.05 to 0.25 m3s-1. In the Petit Glâne (mean Q 0.94 m3s-1), discharge increases are around 0.1 m3s-1, and decreases range from 0.3 to 0.8 m3s-1 (see Appendix ).

The Arbogne resembles the Petit Glâne, and the response of the Broye subcatchment is more attenuated. The share of pervious area per catchment (where SOC is increased) is comparable between the subcatchments: The Broye subcatchment and Arbogne have 61 % and 62 %, Flon and Petit Glâne slightly higher shares with 72 % and 74 % (Fig. ). Thus, the described differences between stations rather arise from climatic variations than differences in land use.

The Flon chatchment has a higher average elevation, with lower temperatures, more precipitation and therefore increased discharge (Fig. ). The Petit Glâne and Arbogne lie lower and receive less precipitation and show therefore also less discharge. The Broye subcatchment spans a wider elevational and climatic gradient, thus slightly averaging out the effects.

Peak flows are, in general, reduced under the SOC scenarios, although the effect is small (Fig. ). Floods with a 2-year return period occurred in winter 2017/2018 and summer 2021 (Q2 events, red vertical lines in Fig. ). Discharge during these events is slightly decreased under the SOC scenarios in 2017/2018, but the impact in 2021 is negligible. For instance, the peak flow in the Broye subcatchment in 2018 reached 77.17m3s-1 and was reduced by 0.2–0.5m3s-1 under the SOC scenarios, while in 2021 a peak of 70.29m3s-1 was reduced by 0.3–1m3s-1 (see Appendix ).

A relevant indicator for low flows is the Q347 threshold, an indicator that cantonal authorities use to determine bans on irrigation water withdrawal from rivers to fulfill the minimum environmental flow requirements. In Fig. , days where the observed discharge fell below this threshold are marked in gray as low-flow days. Although the influence is minor, discharge is slightly increased under the SOC scenarios before and sometimes during observed low-flow periods. This leads to fewer days falling below the Q347 – typically 1–6 d depending on scenario, year, and subcatchment (in the Broye subcatchment, for example, 1–4 d less). However, in the Arbogne in 2016 and 2019, as well as in the Petit Glâne in 2019, low-flow periods coincided with reduced discharge under the SOC scenarios, resulting in more low-flow days (a surplus of 1–14 d in the Arbogne and up to 5 d in the Petit Glâne, Fig. ).

Since the overall small impact of SOC increase on ET was first surprising, we wanted to investigate the responses of the individual soil layers. Here we found, that although the soil water content in the first two layers was consistently higher under the SOC scenarios, ET from soil layers 1 is reduced, while it is increased from soil layer 2 and 3, leading to an overall small net increase. The reason for this is explained and discussed in Sect. .

The SOC scenarios represent possible outcomes of combinations of management adaptations. Their impact on the model output fluxes ET and total grid-scale runoff increases almost linearly with increasing SOC content, as visible by comparing scenarios LowC_vLowC, MedC_LowC and vHighC_MedC in Fig. .

In total, the largest SOC additions occur in scenario vHighC_MedC (+1.5 % in the first layer and +0.6 % in the second layer), whereas in scenario MedC_MedC, SOC is added evenly across both layers (+1 % in each). Interestingly, the effects on soil moisture are often largest under MedC_MedC, despite its slightly lower total SOC increase (Fig. ). This suggests that increasing SOC in the subsoil can be particularly beneficial, as it slows soil moisture depletion during late summer and fall. Differences between vHighC_MedC and MedC_MedC are generally minor: in most catchments, the two scenarios produce nearly identical reductions in low-flow days compared to the base scenario. However, in the Arbogne in 2018, low-flow days are reduced by one day only under MedC_MedC, and in the Flon in 2022, the reduction under scenario MedC_MedC is two days – one day more than under vHighC_MedC.

For the catchment-wide, annual effect, the distribution of SOC in the two soil layers does not make a difference. Only at the seasonal scale, a distribution into deeper layers might lead to a delay of drought-induced transpiration reduction (as was observed in ).

The mHM model is well suited for impact studies like this due to its open-source nature, active user community, and flexible structure, which allows individual adjustments, such as in the estimation of soil hydraulic properties .

As with the case of any modeling scheme, some simplifications and limitations remain. First, in mHM only three land-use classes are distinguished, which may be sufficient at large scales but limits the representation of heterogeneous agricultural landscapes. In our study region, pervious land cover aggregates cropland and meadows, which can differ in management and water-use processes. Introducing additional land-use classes and distinguishing different crop functional types with varying root profiles would improve model realism. Differentiating winter crops and spring crops could be important given their distinct patterns of water uptake which may influence recharge and also low flow dynamics differently.

In the present framework, however, SOC effects are presented primarily through changes in soil hydraulic properties which directly control soil moisture availability and, consequently, ET. While mHM accounts for some vegetation responses (e.g., adjustments in rooting depth; Sect. ), plants are represented at a coarse level.

Regarding ET, mHM separates canopy interception but aggregates soil evaporation and transpiration into a single flux, as is common in many hydrological models . While net ET is likely captured realistically, the partitioning between productive (transpiration) and unproductive (soil evaporation and interception) fluxes, as well as their temporal dynamics, may differ from reality. Finally, the root distribution, which varies with θFC, is more dynamic than standard static profiles, but could be improved by incorporating dynamic crop/root growth and reassessing the negative relationship between θFC and root density under the climatic and edaphic conditions of our study region. Overall, while these limitations can affect plot scale process representation, their impact at the catchment scale is uncertain and difficult to validate, particularly given the lack of direct observations for root distribution or ET partitioning. Within these constraints, representing SOC impacts via soil hydraulic properties captures the dominant hydrological pathway relevant at the catchment scale and is therefore the focus of this study.

Compared to fully physics-based models such as WaSiM-ETH , which often require extensive parameter adjustment and high computational effort, mHM offers a practical balance between spatially explicit process representation and computational efficiency . Fully physics-based models are in practice never “fully” mechanistic, and for our purpose they would not provide additional advantages in representing SOC-related management effects. Their higher data and computational demands would mainly add complexity without improving the core processes relevant to this study.