The study site is located in the Ammer valley floodplain, approximately 7 km west of Tübingen, southwestern Germany (48.521266° N, 8.972687° E, 394 m a.s.l.). The Ammer catchment is situated in the South German Escarpment within geological units of the Middle and Upper Triassic. The western part of the catchment lies within the Muschelkalk unit, whereas the eastern part lies within Keuper units, which mainly consist of terrestrial sediments such as sandstones, mudstones, and dolostones (Geyer et al., 2011; Grathwohl et al., 2013; Martin et al., 2020; Schwientek et al., 2013). The floodplain containing the study site is located in the eastern part of the catchment and is bordered by Keuper hills to the south and north. The hydrostratigraphy of the floodplain comprises, from bottom to top, Keuper rocks as basement, an aquifer mainly composed of gravel and clay (gravel aquifer), a regional aquitard (lower clay aquitard), a confined aquifer composed of calcareous tufa sediments (tufa aquifer), and an upper regional aquitard (upper clay aquitard) (Martin et al., 2020). The investigated soil developed within the upper clay aquitard, which is, on average, 2 m thick and consists of fine-grained, clay-mineral-rich alluvial sediments derived from the surrounding Keuper hills. The lower part of this aquitard is water-saturated throughout the year. The soils are classified as Gleysols and show typical gleyic properties, including visually distinguishable reduced and oxidized horizons (Appendix A; Martin et al., 2020). The 50 cm soil cores collected in this study sampled only the Ah and Go horizons. Although the site is geomorphologically part of the Ammer floodplain, the Ammer currently flows in a regulated riverbed. Regular overbank flooding of the investigated area is therefore unlikely under present conditions. This information is relevant for interpreting microbial responses to drying–rewetting and flooding-related redox fluctuations.

The soil organic matter content, quantified as total organic carbon (TOC; methodological details in Appendix B), was 2 wt %–2.5 wt % in the upper 15 cm and approximately 1 wt % between 15 and 50 cm depth. Soil texture was relatively uniform within the upper 50 cm, with 11 ± 4 wt % sand, 56 ± 6 wt % silt, and 32 ± 4 wt % clay (Appendix C; determined according to DIN EN ISO 17892-4). Soil pH ranged from 7.0 to 7.3 and was determined on six samples from irrigated plots, with two samples each from 0–10, 20–30, and 40–50 cm depth, after stirring the soil in a 1:1 (v:w) mixture with deionized water for 30 min.

The experiment was conducted on arable land after the harvest of summer barley and after a prolonged summer dry period (see Sect. 2.2 for meteorological details). Information on previous land use was provided by the farmer. During winter 2018/2019, before the cultivation of summer barley, a catch crop consisting of radish and mustard was planted and incorporated into the soil with a cultivator in February 2019. The plot was fertilized with horse manure in winter 2018/2019 and with calcium ammonium nitrate fertilizer (27 % N) in spring 2019. During barley cultivation, the herbicide Ariane-C was applied, containing 100 g L⁻¹ fluroxypyr (as methylheptyl ester, 144 g L⁻¹), 80 g L⁻¹ clopyralid, and 2.5 g L⁻¹ florasulam.

The climate at the experimental site is temperate oceanic, with a long-term mean annual precipitation of 718 mm and a mean annual temperature of 9.4 °C for the period 1981–2010 (German Meteorological Service (DWD), 2024). Meteorological data were retrieved from a weather station in Tübingen-Unterjesingen (48.52950° N, 8.98380° E, 439 m a.s.l.), located close to the study site (Agrarmeteorologie Baden-Württemberg, 2021; Appendix D). The experiment was conducted from 16 to 30 July 2019. During the two weeks preceding the start of the experiment, daily temperatures exceeded 20 °C on all but 2 d, and three moderate rainfall events occurred, with 9.1, 5.1, and 33.1 mm precipitation, respectively, according to the classification of the American Meteorological Society (2026).

During the experiment, the minimum night-time temperature was 10.2 °C, with an average night-time temperature of 17.7 °C. The maximum daytime temperature was 36.3 °C, with an average daytime temperature of 24.5 °C. Temperatures exceeded 25 °C on 12 d and 30 °C on 6 d. Following the FAO guideline (Allen et al., 1998), the reference evapotranspiration (ET₀) for July 2019 was calculated as 4.2 mm d⁻¹ (Appendix E). Because the experimental plots had been harvested before the experiment, ET₀ is used here as a meteorological reference for atmospheric evaporative demand rather than as a direct estimate of plot-scale evapotranspiration.

Between 16 and 30 July 2019, two natural rainfall events occurred: one on 21 July from 02:00 to 05:00 local time (2.5 mm), and one heavy rainfall event from 26 July at 20:00 LT to 27 July at 05:00 LT (24.7 mm; Appendix D). During both events, the dry control plots were covered with plastic foil. However, during the natural heavy rainfall event from 26 to 27 July, the plastic foil covering plot D2, where the redox probes were installed, was removed by wind.

The simulated heavy rainfall event rapidly increased soil water storage throughout the sampled 0–50 cm profile, whereas the recurrent moderate irrigation applied thereafter was retained mainly in the upper soil layers (Fig. 3a). On day 0 (T0), 95.6 % of the applied irrigation water was recovered within the 0–50 cm profile. Apparent recovery of the cumulative water input decreased to 50.3 % at T6 and 32.1 % at T14.

Two hours after the simulated heavy rainfall event, 54.2 % of the recovered water was located in the upper 10 cm, while 19.9 % occurred at 10–30 cm depth and 25.9 % had already reached 30–50 cm depth (Fig. 3a). At T6 and T14, more than half of the recovered water was again located in the upper 10 cm. In contrast, only small fractions of the recovered water were found at 30–50 cm depth at T6 and T14 (4.7 % and 14.4 %, respectively). Thus, the initial heavy rainfall event caused rapid deep infiltration, whereas subsequent moderate irrigation contributed mainly to water storage in the topsoil.

The isotope mass balance confirmed rapid redistribution of the D₂O-labeled heavy-rainfall water throughout the sampled soil profile (Fig. 3b). Recovery of the labeled heavy-rainfall water was 108 % at T0, 82 % at T8, and 63 % at T14. Recovery of more than 100 % result from slightly uneven irrigation intensity or confluence of the irrigation water at the shrinkage crack sites where the samples were taken. At T0, 30 % of the recovered labeled water was found in the upper 10 cm, while the remaining labeled water was distributed relatively evenly across the deeper 10 cm depth intervals, each containing 15 %–22 % of the recovered label. At T8 and T14, only 13 % and 4 %, respectively, of the recovered labeled water remained in the upper 10 cm. The relative decrease in labeled water recovery over time was therefore strongest in the topsoil.

In situ redox potentials responded strongly to irrigation and showed pronounced variation with depth and time (Fig. 4). At the start of the experiment, redox conditions were oxidizing throughout the monitored soil profile (> +400 mV vs. SHE). In the irrigated plot, redox potentials declined at all monitored depths after the simulated heavy rainfall event. In the upper soil layers (5–20 cm below ground), redox potentials mostly shifted to reducing conditions within 24 h. In the subsoil (30–50 cm below ground), the decline was slower and more variable; the time required to reach +200 mV ranged from one day at 30 cm depth in sensor 2 to 6 d at 30 cm depth in sensor 1.

After this initial decline, redox potentials at 10, 20, and 30 cm depth showed strong temporal variation and also differed between the two sensors installed in the irrigated plot. Values fluctuated between oxidizing and moderately reducing conditions. At 40–60 cm depth, the decrease in redox potential after irrigation was more gradual and persistent. The strongest short-term fluctuations occurred at 10 cm depth in sensor 1, where redox potentials varied between weakly oxidizing and moderately reducing conditions during the phase of recurrent irrigation. When daily irrigation was interrupted between T10 and T14, redox potentials increased again, reaching weakly reducing to oxidizing conditions even at 30–50 cm depth.

In the dry control plot, redox potentials remained oxidizing throughout most of the experiment (Fig. 4). The only exception occurred between T10 and T14 at 10 and 20 cm depth, when the plastic foil covering plot D2 was removed by wind during a natural heavy rainfall event.

The simulated heavy rainfall event affected ammonium and nitrate masses differently (Fig. 5). Ammonium mass increased only in the uppermost soil layer after rewetting, whereas nitrate mass decreased across the upper part of the profile shortly after irrigation.

Two hours after the simulated heavy rainfall event, ammonium mass in the 0–5 cm layer nearly doubled, increasing from 0.08 ± 0.03 mg N in dry control plots at T0 to 0.15 ± 0.02 mg N in irrigated plots at T0 (p<0.05; Fig. 5a). In the same layer, ammonium mass further increased to 0.20 ± 0.01 mg N at T6 (p=0.058 compared with WT0) and 0.3 ± 0.2 mg N at T14 (p>0.05 compared with WT0). Below 5 cm depth, ammonium masses changed only slightly during the experiment. Consequently, total ammonium mass in the 0–50 cm profile remained similar in irrigated plots across T0, T6, and T14, ranging between 0.8 and 0.9 mg N.

In dry control plots, total ammonium mass increased from 0.8 ± 0.1 mg N at T0 to 1.2 ± 0.3 mg N at T14, although this increase was not statistically significant (p>0.05). The strongest relative increase occurred in the 0–5 cm layer, where ammonium mass increased approximately three-fold (p<0.05). At greater depth, ammonium mass increased by factors of 1.1 to 2.2, but these changes were not significant.

Nitrate showed a contrasting pattern. Total nitrate mass in the 0–50 cm profile decreased by approximately one third within two hours after irrigation, from 1.2 ± 0.1 mg N in dry control plots at T0 to 0.8 ± 0.1 mg N in irrigated plots at T0 (p<0.05; Fig. 5b). This decrease was driven mainly by losses from the upper soil layers. Nitrate mass decreased by approximately 40 % at both 0–5 and 5–10 cm depth (both p<0.05) and by 18 % at 10–20 cm depth (p>0.05).

During the following two weeks, nitrate mass in irrigated plots increased again. Total nitrate mass nearly doubled from 0.8 ± 0.1 mg N at WT0 to 1.5 ± 0.4 mg N at WT14 (p<0.05). From WT0 to WT6, nitrate mass increased mainly in the upper 20 cm, while changes below 20 cm were small. From WT6 to WT14, nitrate mass increased in the 0–5 cm layer and also below 20 cm depth, whereas nitrate masses at 5–10 and 10–20 cm depth decreased; however, these depth-specific changes were not significant.

In dry control plots, total nitrate mass increased from 1.2 ± 0.1 mg N at T0 to 3.0 ± 1.2 mg N at T14, but this increase was not statistically significant (p>0.05). The increase occurred mainly in the upper soil layers, with an approximately three-fold increase at 0–5 cm depth (p>0.05), a 2.4-fold increase at 5–10 cm depth (p<0.1), and a 2.2-fold increase at 10–20 cm depth (p>0.05).

Microbial community composition showed high spatial variability but no clear separation between irrigated and dry control plots over the 14 d experiment. After quality filtering, the 16S rRNA amplicon dataset contained a mean of 7401 reads per sample for DNA and 6970 reads per sample for RNA. Because read numbers varied strongly among RNA samples, only samples with at least 300 reads were retained, reducing the RNA dataset from 54 to 46 samples. The retained samples contained, on average, 1334 amplicon sequence variants (ASVs).

Alpha diversity of the active bacterial community, based on the RNA fraction, did not differ significantly between dry control and irrigated plots (Fig. 6). Diversity indices tended to be slightly higher in dry control plots than in irrigated plots, but this pattern was not statistically significant. Shannon diversity of the active community did not differ between treatments (p=0.956).

Across all samples, the bacterial communities were dominated by Proteobacteria (17.4 ± 4.7 %), Acidobacteriota (14.5 ± 4.9 %), Actinobacteriota (13.5 ± 7.4 %), and Verrucomicrobiota (11.5 ± 5.2 %). These phyla remained dominant irrespective of treatment, depth, and sampling time (Appendix G). At family level, Pedosphaeraceae (Verrucomicrobiota), unclassified Vicinamibacterales (Acidobacteriota), Xiphinematobacteraceae (Verrucomicrobiota), and Nitrosomonadaceae (Gammaproteobacteria) were the most abundant groups, with mean relative abundances ranging from 3.2 % to 4.6 %.

Ordination of DNA- and RNA-based community composition showed no clear clustering by treatment or soil depth (Fig. 7). This indicates that neither total nor active bacterial community composition changed consistently in response to irrigation over the 14 d observation period.

In contrast to community composition, extracellular enzyme activities responded clearly to irrigation and soil depth (Fig. 8). Activities of β-glucosidase, exochitinase, and acid phosphatase all decreased significantly with soil depth (p < 0.05). β-glucosidase and exochitinase activities were significantly higher in irrigated plots than in dry control plots (p < 0.05), whereas acid phosphatase activity did not differ between treatments (p > 0.05). Sampling time after rewetting was not a significant factor for any of the three enzymes (p > 0.05). The stimulation of β-glucosidase and exochitinase by irrigation was strongest in the upper soil layer and declined with depth, as reflected by significant treatment × depth interactions for both enzymes (p < 0.05).

The combined water and isotope mass balances showed that the simulated heavy rainfall event caused rapid infiltration throughout the sampled 0–50 cm profile. Two hours after irrigation, only about half of the recovered water was retained in the upper 20 cm, whereas a substantial fraction had already reached deeper soil layers (Fig. 3a). The δ2H isotope mass balance confirmed that D₂O-labeled heavy-rainfall water was distributed across the entire sampled profile shortly after irrigation (Fig. 3b). This rapid deep infiltration is consistent with preferential flow through the pronounced shrinkage-crack network that formed in the fine-textured soil during the preceding drought period (Schlögl et al., 2022).

The hydrological response changed markedly after the initial wetting. During the subsequent phase of recurrent moderate irrigation, the apparent recovery of total water input within the sampled profile decreased from 96 % at T0 to 50 % at T6 and 32 % at T14 (Fig. 3a). At the same time, newly added water was retained mainly in the upper soil layers. This pattern indicates that the initial heavy rainfall event used open shrinkage cracks as preferential flow paths, whereas later moderate irrigation occurred after crack closure and therefore infiltrated more slowly into the soil matrix. Because the plots had been harvested before the experiment, water losses were likely driven mainly by evaporation and drainage below the sampled profile rather than by plant transpiration. The calculated ET₀ of 4.2 mm d⁻¹ therefore represents atmospheric evaporative demand rather than plot-scale evapotranspiration. The stronger decline in labeled-water recovery from the uppermost 10 cm compared with deeper layers further supports strong near-surface water loss after the initial event.

Nitrate and ammonium responded differently to this hydrological redistribution. Nitrate mass decreased strongly in the upper soil layers within two hours after the simulated heavy rainfall event, whereas ammonium mass did not show comparable profile-level losses (Fig. 5). This contrast is consistent with the different mobility of the two inorganic nitrogen species. Nitrate is weakly sorbed and can be transported rapidly with percolating and infiltrating water, whereas ammonium is retained more strongly by cation exchange at negatively charged mineral surfaces, particularly clay minerals (Blume et al., 2016). Thus, preferential flow after drought likely caused rapid vertical nitrate displacement, while ammonium remained largely retained in the sampled soil profile.

The decrease in nitrate mass in the upper soil profile was not accompanied by a corresponding nitrate accumulation at 30–50 cm depth. This suggests that at least part of the displaced nitrate may have moved below the sampled 50 cm profile. Alternative explanations, such as denitrification or plant uptake, are less likely to fully explain the immediate nitrate decrease. Redox potentials were still oxidizing shortly after the simulated heavy rainfall event, and crops had been harvested before the experiment. Microbial immobilization may have contributed, but the rapid timing and magnitude of the nitrate decrease are more consistent with hydrological export along preferential flow paths. Nevertheless, because nitrate was not measured below 50 cm depth, nitrate transfer beyond the sampled profile should be interpreted as a likely but indirect inference.

The potential transfer of nitrate below the biologically active topsoil is relevant for groundwater quality, particularly in fine-textured floodplain soils, where shrinkage cracks can bypass the soil matrix. The study site is underlain by a shallow tufa aquifer below the upper clay-rich layer (Martin et al., 2020). If nitrate reaches such deeper zones repeatedly, its fate will depend on the local redox status and denitrification capacity of the subsurface. Therefore, heavy rainfall following drought may represent a short but important hydrological window during which nitrate can bypass retention and transformation in the upper soil profile.

The increase in ammonium mass in the uppermost soil layer shortly after irrigation is consistent with a rewetting-induced mineralization pulse, commonly referred to as the Birch effect. Ammonium mass nearly doubled in the 0–5 cm layer within two hours after the simulated heavy rainfall event (Fig. 5a). This response agrees with earlier observations that rewetting of dry soils can cause a transient release of inorganic nitrogen through microbial cell lysis, mineralization of dead microbial biomass, and decomposition of osmoregulatory compounds released during drought and rewetting (Birch, 1958, 1960; Fierer and Schimel, 2002; Jarvis et al., 2007; Saetre and Stark, 2005; Unger et al., 2010). The simultaneous increase in β-glucosidase and exochitinase activities supports the interpretation that rewetting rapidly stimulated microbial and enzymatic activity in the topsoil. The increase in ammonium mass cannot solely be caused by atmospheric deposition (Appendix I). This mineral nitrogen pulse necessarily implies mobilization and mineralization of organic nitrogen, potentially including labile soil organic nitrogen and nitrogen released from drought-stressed or lysed microbial biomass. Drying–rewetting can increase the availability of such organic nitrogen substrates and stimulate net nitrogen mineralization after rewetting, although the contribution of individual organic nitrogen pools cannot be quantified from the present measurements (Maxwell et al., 2022; Schimel, 2018).

Below 5 cm depth, ammonium masses changed only weakly in irrigated plots. Three, non-exclusive mechanisms may explain this limited response. First, labile organic matter decreased with depth (Appendix B), reducing the potential for rewetting-induced ammonium release. Second, ammonium may have been retained by sorption to clay minerals and partly fixed in 2:1 clay interlayers, reducing its mobility and microbial accessibility (Buss et al., 2004; Cavalli et al., 2015). Third, newly produced ammonium may have been balanced by microbial uptake or oxidation under locally oxic conditions. Because gross nitrogen transformation rates were not measured, these mechanisms cannot be separated quantitatively.

Nitrate dynamics after the initial rainfall event were more complex. After the immediate decline in nitrate mass at T0, total nitrate mass in irrigated plots increased again by T14 (Fig. 5b). This increase cannot be fully explained by atmospheric nitrate deposition, which was small relative to the observed nitrate changes (Appendix J). Periods with higher redox potentials may have favored nitrification, whereas lower redox potentials may have favored nitrate reduction. The redox records support this interpretation qualitatively: redox potentials declined after irrigation but showed strong temporal variation, particularly at 10–30 cm depth, and increased again during the interruption of daily irrigation between T10 and T14 (Fig. 4).

However, these data do not allow direct quantification of nitrification, denitrification, immobilization, or mineralization rates. Therefore, we interpret the observed nitrate patterns as redox-associated changes in inorganic nitrogen pools, not as direct evidence for a process-level switch in nitrogen cycling. This distinction is important because nitrate and ammonium pool sizes integrate multiple simultaneous processes, including production, consumption, immobilization, and transport. In addition, redox potentials were measured with high temporal resolution but only in one irrigated and one dry control plot. The sometimes very marked differences in the sensors' responses can be caused, for example, by the highly uneven distribution of dry cracks or remaining root fragments from the harvested plants. The redox data therefore provide valuable mechanistic context, but spatial extrapolation to all replicate plots must remain cautious.

The strong spatial and temporal variability in nitrate and ammonium masses further highlights the difficulty of inferring nitrogen process rates from pool-size data alone. Coefficients of variation (Appendix H) showed that inorganic nitrogen distributions did not simply mirror the spatial pattern of the D₂O-labeled water. While preferential flow clearly governed rapid water redistribution during the initial event, nitrate and ammonium masses were also shaped by sorption, biological transformations, and local soil heterogeneity. Thus, drought-induced cracks were important for rapid water and nitrate displacement, but they were not the only control on inorganic nitrogen pool distributions during the subsequent 14 d.

The increase in ammonium and nitrate masses in dry control plots between T0 and T14 also illustrates this interpretive complexity. Because these plots remained dry for longer, continued drought may have intensified microbial stress and contributed to organic matter and microbial biomass turnover, with subsequent accumulation of inorganic nitrogen. As atmospheric nitrogen inputs were too small to account for this increase, part of the additional inorganic N must ultimately have originated from net mineralization of soil organic nitrogen and/or microbial biomass nitrogen (Appel, 1998). The simultaneous accumulation of NH4+ and particularly NO3- is therefore consistent with continued organic nitrogen mineralization followed by nitrification, although neither process can be quantified from the observed pool changes alone. In addition, evaporation may have promoted upward redistribution and concentration of dissolved nitrate in the upper soil layers. The temporary plastic cover used to exclude natural rainfall may also have altered temperature and moisture conditions near the soil surface. These explanations remain tentative, but they emphasize that inorganic nitrogen pools can change substantially even without experimental irrigation.

Overall, the nitrogen data support two main conclusions. First, rewetting caused a rapid but spatially restricted ammonium increase in the topsoil, consistent with a short-term Birch-effect response. Second, nitrate pools after the initial event were associated with changing hydrological and redox conditions, but the underlying process rates cannot be resolved without direct measurements of gross mineralization, nitrification, immobilization, and denitrification.

Microbial responses to drought and rewetting differed between community composition and extracellular enzyme activity. Bacterial community composition showed no consistent treatment-related separation at either the DNA or RNA level during the 14 d experiment (Fig. 7). Alpha diversity of the active community also did not differ significantly between dry control and irrigated plots (Fig. 6). These results indicate that the short-term rewetting event did not cause a detectable restructuring of bacterial community composition at the taxonomic resolution captured by 16S rRNA amplicon sequencing.

The absence of a clear compositional response does not imply that microorganisms were inactive. In contrast, β-glucosidase and exochitinase activities increased significantly in irrigated plots, especially in the upper soil layer (Fig. 8). This indicates that rewetting rapidly affected microbial activity or extracellular enzyme availability, even though community composition remained comparatively stable. Such a decoupling between microbial activity and community composition is plausible because enzyme activities can respond within hours to changes in water availability, substrate diffusion, cell lysis, and microbial metabolic activation, whereas detectable shifts in community composition may require longer time scales or stronger selective pressure.

Several factors may explain why microbial community composition remained stable. First, floodplain and arable soils are highly heterogeneous at small spatial scales, and such micro-scale variability can mask treatment effects in community ordinations (Vos et al., 2013). Second, DNA can persist in soils as relic DNA or in dormant cells, which may obscure short-term changes in the living community (Carini et al., 2016; Lennon et al., 2018). Although RNA was analyzed as a proxy for the active community, rRNA-based community profiles may still reflect both activity and standing ribosome pools rather than instantaneous growth or process rates. Third, the 14 d observation period may have been too short to detect consistent taxonomic restructuring, particularly if the community response was mainly physiological.

The slightly higher RNA-based diversity indices in dry control plots than in irrigated plots were not statistically significant and should therefore not be interpreted as evidence for selection of better-adapted taxa. Rewetting can impose osmotic stress and can eliminate sensitive cells, while also stimulating dormant or rare taxa (Aanderud et al., 2015; Barnard et al., 2013; Blazewicz et al., 2014; Schimel, 2018). However, our data provide stronger evidence for an activity response than for compositional selection. Therefore, the microbial results are best interpreted as showing rapid functional activation after rewetting without detectable community restructuring over the short experimental period.

This interpretation is consistent with the enzyme and nitrogen results. Increased β-glucosidase and exochitinase activities after irrigation indicate enhanced potential for C and N acquisition from organic substrates and support the observed short-term ammonium increase in the topsoil. In contrast, the lack of a clear microbial community shift suggests that the immediate biogeochemical response was driven primarily by altered microbial activity, substrate accessibility, and physical–chemical conditions rather than by major changes in bacterial community composition.