The soil used in the study was collected from the “La Cage” field experiment on the French National Institute for Agricultural Research (INRAE) campus in Versailles, Ile-de-France (48°45^′ N; 2°08^′ E). The climate of the greater Parisian basin is characterized as temperate, with an average precipitation of 630 mm and an annual average temperature of 10.4 °C (Bellone et al., 2023). The experimental site was established in 1998 to conduct long-term field studies on the effects of different cultivation methods on a well-drained Luvisol soil (Autret et al., 2020). Soil from two treatments, conventional tillage and conservation agriculture, were tested. The conventional tillage site underwent annual ploughing to a depth of 30 cm and systematic pesticide application. In contrast, the conservation agriculture treatment was not tilled and was maintained with persistent cover crops. Pesticides were applied only when plant damage was detected in the conservation agriculture treatment.

Both fields were under plant cover at the time of sampling in early November 2023. This cover corresponded to mustard in the case of the conventional tillage field, and common vetch and black oats for the conservation agriculture field. The main crops consisted of a wheat-pea-rapeseed crop rotation for the conventional tillage site and pea-wheat-corn-oat rotation for the conservation tillage site. A more detailed description of the plant cover and management practices is provided in Autret et al. (2016). The choice to use the conventional tillage and conservation agriculture treatments was based on the expectation that these would have different soil structures and C contents, both key factors influencing carbon dynamics. The selection of these two systems is supported by findings from the La Cage study, which show that conservation agriculture generally sustains higher microbial biomass, total C, and significantly greater stabilized soil organic carbon content compared to conventional tillage (Henneron et al., 2015; Juarez et al., 2013). Further studies have identified the presence of more stable water-soluble aggregates and associated improved porosity in the conservation agriculture fields compared to conventional tillage treatments (Chabert and Sarthou, 2020; Cosentino et al., 2006). This improved aggregate stability is expected to influence the maintenance of pore connectedness across a range of matric potentials (Mondal and Chakraborty, 2022).

To obtain a wide range of pore size distributions, pore connectivity, carbon contents and microbial biomass, we used sieved (< 2 mm) and undisturbed samples from the topsoil (0–15 cm) and subsoil (50 cm). The soil microbial biomass and soil C contents are known to decrease with depth of the soil profile (Salomé et al., 2010) and sieving changes the physical structure and water retention properties of soil (Herbst et al., 2016).

The undisturbed samples were collected by inserting PVC sampling cores (80 mm in length and 58 mm in internal diameter) directly into the soil horizons and carefully extracting them to preserve soil structure. All undisturbed samples were frozen at -20 °C until use to minimize organic matter decomposition during storage prior to incubation experiments (Allende-Montalbán et al., 2024). Bulk soil collected to prepare sieved samples was air-dried, sieved to 2 mm, and stored at 10 °C until the experiment began.

Undisturbed samples were thawed gradually by transferring them from a deep freezer to an 8 °C refrigerator for 24 h, followed by five days at 20 °C in the laboratory prior to incubation. Sieved samples were prepared by packing soil to a target bulk density of 1.73 g cm⁻³, corresponding to the average bulk density measured in undisturbed samples. Because the undisturbed cores were collected under very wet field conditions, some disturbance during sampling may have affected the measured bulk density. This bulk density value was therefore used only to obtain comparable soil mass and pore volume between treatments and not to reproduce the in situ soil structure.

Samples were saturated by capillary rise by initially placing them in 1 cm of water, with additional water added gradually over 24 h until full saturation was achieved. The samples were then transferred to airtight microcosms and incubated in the dark to limit algal growth on the microcosm surface caused by light exposure.

This study examined the effect of soil water status on respiration rates and electrical conductivity during short-term incubations. To impose different soil water contents without physically disturbing the samples, matric potentials of -70, -100, -250, -350, -450 and -630 hPa were applied using a suction system. The corresponding conversion between matric potentials and water content is described in Sect. 2.4.3.

Samples were placed in airtight microcosms fitted with a ceramic plate allowing adjustment of the pressure head and a septum for headspace sampling (Poll et al., 2010; Fig. 1). The ceramic plate positioned at the base of each microcosm was connected to a suction pump, enabling precise pressure-head control of the soil cores. Additionally, an opening in the lid was used to route electrical cables for conductivity measurements. A schematic of the setup is provided in Fig. S1 in the Supplement.

Five microcosms at a time were subjected to controlled suction, and drained water was collected in refrigerated bottles at 4 °C for subsequent analysis of soil solution ionic strength. One microcosm was used exclusively for gravimetric measurements to verify changes in moisture content. Each treatment was performed in triplicate.

Incubations were conducted at 20 °C. Samples were maintained at each pressure head for approximately 24 h after leaching had ceased to allow hydraulic equilibrium to be reached. In total, 24 microcosms were used (four soils × two structural treatments × three replicates). Samples were analysed in randomized order such that replicates from the same soil and treatment combination were not measured simultaneously.

After measurements at -630 hPa were completed, samples were transferred to mason jars fitted with septa for headspace sampling and containing an oversaturated lithium chloride solution. This induced a suction at the sample surface corresponding to a pressure head of -996 hPa (Colas, 2011; Greenspan, 1977). Sample weights were monitored periodically until no further mass loss was observed, indicating that the target pressure head had been reached.

Organic matter (OM) content varied among individual soil samples (Fig. S3), ranging from approximately 3.6 % to 14.1 % of dry soil mass. The lowest OM content was observed in the sieved CT topsoil, whereas the highest values occurred in the undisturbed CT subsoil samples. However, OM content did not differ significantly among groups (Fstatistic(7,10.4)= 1.96, p= 0.16). At the highest saturation level, a significant difference in total respiration rate was observed between sieved and undisturbed samples (p < 0.03), as well as between topsoil and subsoil samples (p< 0.002) (Table S2).

The relative respiration rate, normalized to the values measured at -70 hPa, increased by approximately three orders of magnitude as pressure head increased (i.e., became less negative) from -996 to -70 hPa (Fig. 1). For the sieved samples, this trend was maintained across all pressure heads, with a small respiration peak observed at -100 hPa. In contrast, the undisturbed samples exhibited a respiration peak at -250 hPa. This increase was particularly pronounced for the topsoil conservation agriculture (CA) sample.

Apart from this peak in the undisturbed samples, no significant differences were observed between sampling depths (topsoil versus subsoil) or field treatments (CT versus CA). Linear mixed-effects modelling indicated a significant effect of pressure head on respiration flux (p= 0.0008), but no significant effects of sample structure (p= 0.178), field treatment (p= 0.436), or their interactions.

In addition, no significant difference in relative respiration flux was detected between -996 and -630 hPa (p>0.618). Post hoc comparisons further confirmed that relative respiration flux did not differ significantly among pressure heads beyond -450 hPa (p= 0.311).

Electrical conductivity (EC) measurements did not differ significantly between electrode channels within samples. Higher values were observed in the sieved conventional tillage (CT) topsoil (1052 µS cm⁻¹) and subsoil (742 µS cm⁻¹). Results for the conventional tillage (CT) samples are shown in Fig. 2, while the corresponding results for CA soils are provided in Fig. S5.

Across both CT and conservation agriculture (CA) systems, EC increased systematically with increasing water saturation (Figs. 2 and S5). Although variability among individual cores was observed, no consistent separation between undisturbed and sieved treatments was evident across the full saturation range. Likewise, differences between topsoil and subsoil samples were not systematic. Two-way ANOVA applied to the fitted Archie saturation exponent (n, log-transformed) confirmed the absence of significant effects of depth, soil structure, or their interaction in either system (CT: depth p= 0.72, structure p= 0.26, interaction p= 0.68; CA: depth p= 0.89, structure p= 0.77, interaction p= 0.90).

In the undisturbed CT treatment, one replicate in both topsoil and subsoil showed a steeper decline in EC with decreasing water saturation (Fig. 2), leading to greater variability in the fitted EC–Sw relationships. This pattern was not observed in the CA samples or in other depth-structure combinations.

The EC-derived tortuosity factor (τw) increased with decreasing water saturation for all samples (Figs. 2 and S5), reflecting progressively reduced diffusive transport at lower saturation levels. Although undisturbed samples exhibited greater variability across the saturation range, two-way ANOVA applied to log-transformed τw revealed no significant effects of depth, soil structure, or their interaction in either management system (CT: depth p= 0.57, structure p= 0.22, interaction p= 0.55; CA: depth p= 0.62, structure p= 0.56, interaction p= 0.64). Similar results were obtained when τw values derived from the fitted curves were compared at Sw= 0.3, 0.5, and 0.8 for both management systems.

To evaluate whether respiration flux was coupled to diffusive transport constraints, linear regressions were fitted on log₁₀-transformed respiration flux and log₁₀-transformed EC-derived tortuosity (τw) for both CT and CA samples (Figs. 3 and S6 for CA).

In the CT samples, respiration flux was negatively related to τw. Linear regressions were fitted using individual measurements from each soil core across the range of water saturations. A linear model including depth and structure confirmed a significant interaction between τw and soil structure (p< 0.001), indicating that slopes differed between undisturbed and sieved samples. No significant interaction with depth was observed. Slopes and coefficients of determination for each depth-structure combination are summarised in Table S6.

In the CA samples, a significant overall negative relationship between respiration flux and τw was also observed (Fstatistic(1,58)= 5.08, p= 0.028). However, neither depth (p= 0.90) nor soil structure (p= 0.72), nor their interactions with τw significantly influenced the slope of the relationship. These results indicate that respiration decreased with increasing tortuosity in both management systems, although structural effects were only detectable in the CT samples.

We hypothesised that decreasing matric potential would be associated with reductions in both CO₂ flux and EC, reflecting reduced aqueous phase connectivity within the pore network (Fig. 3). Reduced water availability is known to increase hydraulic and electrical tortuosity, thereby constraining diffusive transport (Ghanbarian et al., 2013; Revil and Jougnot, 2008). Consistent with this conceptual framework, both EC and soil respiration decreased as the soils dried. These declines were accompanied by an increase in τw with decreasing water saturation (Fig. 2).

Respiration peaked at -250 hPa in the undisturbed samples and did not increase further at higher saturation levels. Because respiration near saturation may reflect oxygen limitation rather than substrate diffusion, the highest saturation levels were excluded from the regression analyses linking respiration and tortuosity (Fig. 3). This plateau likely reflects reduced oxygen diffusion under near-saturated conditions, as oxygen diffuses more slowly through water than through air (Moyano et al., 2013). Only measurements within the diffusion-limited range below this peak were therefore included in the regression analyses, and this exclusion was applied consistently across structural treatments to ensure comparability.

An increase in respiration flux was observed at -250 hPa in the undisturbed samples. At higher saturation levels (-250 to 0 hPa in undisturbed samples and -100 to 0 hPa in sieved samples), respiration did not increase further. This plateau may reflect reduced oxygen diffusion under near-saturated conditions, as oxygen diffuses more slowly through water than through air (Moyano et al., 2013). Similar non-monotonic responses of respiration to water content have been reported across soil types (Ebrahimi and Or, 2015; Moyano et al., 2013), suggesting the presence of a moisture range where oxygen supply and substrate diffusion are balanced.

Below this moisture range, respiration declined with decreasing water content. At lower saturation, reduced water availability increases tortuosity and constrained diffusive transport (Ebrahimi and Or, 2015; Ghezzehei et al., 2019). In this regime, EC and the derived tortuosity factor (τw) reflect changes in the geometry of the conductive phase and therefore provide information on conditions affecting diffusive transport (Revil and Jougnot, 2008; Jougnot et al., 2009).

A general decrease in EC with decreasing water saturation was observed in most samples (Fig. 2). This pattern is consistent with reduced connectivity of the water phase within the pore system, which increases electrical tortuosity and lengthens transport pathways (Ghanbarian et al., 2013; Jougnot et al., 2018). At sufficiently low saturation, the water phase approaches a percolation threshold where conductive pathways become increasingly disconnected (Hunt, 2005; Revil and Jougnot, 2008). Because solute and substrate diffusion occur primarily through the aqueous phase, this progressive loss of connectivity supports the interpretation that decreasing water saturation constrains microbial respiration through diffusion limitation.

Interestingly, at the second lowest water content in the conventional tillage subsoil samples, EC showed an unexpected increase. A clear physical explanation is not evident; however, fungal growth was observed on the sample surface during the incubation period. Although speculative, fungal activity may have contributed to this isolated increase in EC. Fungal hyphae can modify their surrounding microenvironment through both chemical and physical processes, including mobilisation of phosphate ions and aggregate formation (Ameen et al., 2019; Hartmann and Six, 2022; Sun et al., 2022). Such local changes in pore-water chemistry or structure could influence measured conductivity. This feature was not observed in other depths or management systems, and therefore remains tentative.

In several cases, respiration did not differ significantly between the two lowest water saturations (Fig. 1). This is consistent with reports of sigmoidal declines in respiration with decreasing water availability, where reductions may be gradual across specific matric potential ranges (Curiel Yuste et al., 2007; Ghezzehei et al., 2019). A conceptual summary of the proposed mechanisms is shown in Fig. 4.

The significant log-log relationship between respiration flux and EC-derived tortuosity (τw) observed within the diffusion-limited range (Fig. 3) indicates a coupling between microbial respiration and diffusive transport constraints. In the CT samples, steeper slopes were observed for sieved soils relative to undisturbed soils, suggesting that under homogenised structural conditions respiration responded more strongly to changes in tortuosity. This contrast was not evident in the CA samples.

We hypothesized that sieving would alter pore structure by producing a more homogeneous pore network, leading to a more uniform decline in EC and respiration compared to undisturbed samples. However, while sieving significantly affected respiration rates, this was not reflected in bulk EC values. Instead, the EC-derived tortuosity (τw) showed a more homogeneous response to water saturation in the sieved samples (Fig. 2d and h), consistent with a more uniform pore network following soil homogenisation. Respiration rates were also higher in sieved samples near saturation (Fig. 1). Our findings align with Herbst et al. (2016) who reported similar contrasts between sieved and undisturbed cores across a range of water contents.

Two mechanisms may explain the differences between sieved and undisturbed samples. First, sieving alters pore size distribution, modifying hydraulic connectivity and drainage behaviour (Castellini and Ventrella, 2012; Shaxson, 2006). Smaller and more uniformly distributed pores may retain water longer during drying, reinforcing contrasts between structural treatments at intermediate matric potentials (Çelik et al., 2021; Gupta Choudhury et al., 2014).

Second, sieving can affect carbon availability to microbial decomposers. The physical protection of organic C from microbial decomposition is believed to be a major mechanism leading to organic C persisting in soil (Moyano et al., 2013; Six et al., 2006). Sieving may disrupt this protection by breaking aggregates and increasing contact between substrates and microbial communities. Improved aeration and substrate availability in sieved samples may therefore explain why soil respiration remains elevated even close to complete saturation. The subsequent decline in respiration beyond the optimal matric potential further supports moisture limitation of decomposition (Crawford, 2005; Fan et al., 2015). The absence of significant differences in EC between structural treatments indicates that sieving did not substantially alter bulk conductive properties. This difference in slopes indicates that soil structure influenced the sensitivity of respiration to tortuosity, even though τw itself did not differ significantly between structural treatments. Such decoupling between bulk physical transport metrics and biological response has previously been reported when bulk density and water content are comparable (Nadler, 1991).

Although there were no significant differences in either relative respiration rate or relative electrical conductivity between the management systems, differences emerged in how respiration responded to changes in water availability and tortuosity. In particular, undisturbed topsoil samples from the CT system maintained comparatively higher respiration rates at higher matric potentials relative to the corresponding CA samples. Conventional tillage disrupts soil aggregates and redistributes organic matter within the plough layer, potentially enhancing substrate accessibility and microbial activity in topsoil horizons (Franco-Luesma et al., 2020; Zsolt et al., 2020).The stronger structural effects observed in the CT samples, particularly in the respiration–τw relationship, are consistent with this mechanism, as respiration in sieved CT soils responded more strongly to changes in tortuosity than in undisturbed soils (Fig. 3). In contrast, the absence of comparable structural modulation in the CA samples suggests that long-term aggregate stability may buffer respiration responses to changes in tortuosity.

We hypothesized that respiration–EC relationships would be weaker in the topsoil than in the subsoil due to a more homogeneous distribution of organic matter and therefore reduced diffusion dependence. However, no significant differences were detected. Because organic matter contents were similar between depths, substrate availability was likely comparable, and depth-related differences in diffusion dependence could not be resolved.

Nevertheless, respiration rates in the topsoil were higher than in the subsoil at high moisture levels, indicating greater mineralisation potential under near-optimal conditions. As the soils dried, the respiration rates in the topsoil and subsoil converged, suggesting that diffusion limitation increasingly constrained activity and masked depth-related differences. In sieved samples, this convergence between depths was less apparent, possibly because homogenisation reduced structural and substrate heterogeneity, thereby altering microbial access to organic matter (Salomé et al., 2010).