The study area is located in a recently restored area of the Traisen River (48°22.386′ N, 15°49.613′ E; LIFE+ Traisen project, www.life-traisen.at), one of the Danube tributaries in Austria. The Traisen became heavily modified after the construction of the Altenwörth hydropower plant in 1976. A river stretch close to the mouth with the Danube was extended by 7.5 km, straightened, and regulated by flood protection dams. These modifications reduced the hydrologic connectivity between river and floodplain, resulting in a loss of natural aquatic and terrestrial habitats.

For the experiments, soil and sediment cores were collected along an inundation gradient consisting of three sites: a permanently inundated site (river sediments), an intermittently inundated site (parafluvial sediments), and a rarely-inundated site (floodplain soil). At each sampling site, 6 intact cores (h = 10–15 cm, d = 14 cm) were collected using an iron corer (d = 14 cm) in a total of 18 cores (6 cores per 3 sites). Immediately after sampling, the intact cores were transferred to the mesocosms apparatus used for the incubations (Figure 1) and then transported to the laboratory.

Intact soil and sediment cores were subject to two consecutive drying-rewetting cycles in laboratory experiments (Figure 2). The ¹⁵N isotope technique was used to quantify the contribution of nitrification and denitrification to N₂O emissions in response to drying-rewetting cycles.

Plexiglas mesocosms (485 mm height, 142 mm diameter, covered with black PVC foil) were used for the intact core incubations (for more detailed information on the cores, see Figure 1). For the ¹⁵N-label application, all cores were inundated (total of 4 days), following an initial incubation settling period. The overlying water was enriched with ¹⁵NO3- to a target value of 110 μmol L⁻¹ (using K¹⁵NO₃ [98 ¹⁵N atom%] Sigma-Aldrich). A water pump was placed in each mesocosm to ensure complete mixing in the mesocosms and avoid stratification of the water column. In situ water was used in all cores.

After this period, a total of nine cores (three cores per three sites) were kept inundated until the end of the experiment (long inundation cores; control). The remaining nine cores (drying-short inundation cores; treatment) were subject to two consecutive drying-rewetting cycles, Cycle 1(C1) and Cycle 2 (C2), respectively. Each cycle consisted of a drying period (14 days) followed by a short inundation period (4 days) (Figure 2). At the end of each short inundation period, the cores were allowed to drain completely from the bottom, mimicking percolation in natural conditions where the top layer that becomes exposed to the atmosphere after inundation. The length of the cycles was set considering N dynamics and enhancement during the drying period [first 10–14 days; e.g., (19)], and increases in microbial activity due to a rapid substrate release and lysis of bacterial cells in the first few days after rewetting [~4 days; e.g., (15)].

Samples of soil and sediment and N₂O were taken from all cores at fixed sampling dates in each hydrological cycle (C1 and C2) accounting for the initial and final of each period: on day 1 and day 14 during the drying period, on day 1 and day 4 during the short inundation period (Figure 2). The long inundation cores were equally sampled at the same dates to be compared with the dry-wet effects.

The analysis was carried out in the R environment for statistical computing (R version 4.0.; RStudio version 1.2.1335, RStudio, Inc.). A redundancy analysis (RDA) was applied to relate N₂O fluxes to environmental parameters and determine which explanatory variables best explain the variation in the fluxes (46). The type II scaling was used as to analyze the correlative relationships between variables. N₂O fluxes from nitrification and denitrification were entered as species data, and mineral N (NH4+, NO3-, and NO2-), DOC and OM quantity and quality (PARAFAC components, BIX, HIX, FI) were entered in the RDA as explanatory variables. The categorical variables inundation gradient (site), frequency (cycle number), and phase (drying and inundation) were also included as explanatory variables. A forward selection procedure (1,000 permutations) was used to select statistically significant explanatory variables. The resulting model was tested with an ANOVA-like permutation test for RDA (PERMANOVA, 999 permutations). All the aforementioned statistical analyses were computed with the vegan package (47). Post-hoc comparison between groups was performed after a significant result determined by PERMANOVA, using the non-parametric multivariate analog of the t-statistic, with significance determined by permutation [pairwiseAdonis package; (48)]. To compare the magnitude of the difference between the group means, we computed Cohen's d (or standardized differences), defined as the difference between the means divided by the root mean square of two standard deviations (49) [rstatix package; (50)]. Larger d-values reflect greater effect sizes and positive d values indicate higher flux values.

Initial N-NO3- concentrations are higher in the floodplain soil and decrease along the inundation gradient, with lower concentrations in river sediments for both control and treatment cores (Supplementary Table 0). A similar trend observed for N-NO2- concentrations. Initial N-NH4+ concentrations were lower in the parafluvial sediments (control and treatment cores), and higher in the floodplain soil (control cores) and river sediments (test cores).

Drying significantly increased N-NO3- concentrations compared to long and short inundation at all sites (Figure 3; Supplementary Tables 1, 2). The floodplain soil had significantly higher N-NO3- concentrations than the parafluvial and river sediments during cycle 1 (all phases; Supplementary Tables 1, 2) and cycle 2 (drying phase; Supplementary Tables 1, 2). Similarly, N-NO2- concentrations tended to be higher during the drying phase at all sites (Figure 3) and were significantly higher in the floodplain soil than in the parafluvial and river sediments for both cycle 1 and cycle 2 (all phases; Supplementary Tables 1, 2). N-NH4+ concentrations tended to be higher during the drying phases in the river sediments rather than in the floodplain soil and parafluvial sediments, while the opposite was observed during the inundation phases (Figure 3; Supplementary Tables 1, 2).

DOC concentration was significantly higher in the floodplain soil than in the other sites (both cycles) and tended to be lower during the short inundation phase in the floodplain soil and river sediments (Figure 4; Supplementary Tables 3, 4). OM content (Figure 4) was significantly higher in the floodplain soil than in the river sediments (both cycles; Supplementary Tables 3, 4) and parafluvial sediments (cycle 2; Supplementary Tables 3, 4).

The higher HIX values indicate a higher degree of humification in the floodplain soil extracts than in the parafluvial and river sediment extracts (all phases). The parafluvial and river sediments present a higher fresh-like to humic-like ratio (BIX) than the floodplain soil. The FI ranges for all sites indicate that DOM is from allochthonous sources (average FI values ~ 1.2). FI values are generally higher in the drying phase than in the short inundation phase in all sites and for both cycles (Figure 5; Supplementary Table 5).

The PARAFAC analysis identified 3 components consistent with those previously reported and well-matched with the OpenFluor database (http://www.openfluor.org; Supplementary Table 6). The excitation-emissions regions of each component can be found in Supplementary Figure 1. The modeled component 1 (C1) is described as a terrestrial, humic-like fluorophore originating from degradation of plant and/or soil derived DOM. Component 2 (C2) is also described as a terrestrial humic-like fluorophore derived from lignin (medium processing) and microbial activity. Component 3 (C3) is described as a labile protein-like fluorophore (tryptophan-like) of microbial origin and recent biological production. The R.U. (components 1–3) tended to decrease from the drying phase to the short inundation phase and increased from the short inundation phase to the drying phase at all sites, with less variation between phases for the parafluvial sediments (Supplementary Table 7). For all treatments, the relative proportion of component 1 and 2 was higher in the floodplain soil (C1% and C2%: ~40 and ~50%, respectively), than in the parafluvial and river sediments (C1% and C2%: ~30 and ~40%, respectively), and component 3 was higher for the parafluvial and river sediments (C3%: ~30%) than in the floodplain soil (C3%: ~10%; Supplementary Table 7).

Despite the typically high variability of N₂O emissions in both long inundation cores and drying-rewetting cores, there was evidence of higher N₂O emissions during the drying phase of cycle 2 (Figure 6; Supplementary Table 8). Comparing both phases, N₂O fluxes derived from nitrification were significantly higher in the drying phase than in the long inundation treatment (F = 20.32, p = 0.003) in the river sediments during cycle 1. Cycle 2 showed more distinct patterns. N₂O fluxes from nitrification were significantly higher under the drying phase vs. short inundation phase (F = 3.25, p = 0.012) in the river sediments. N₂O fluxes from denitrification were also significantly higher in the drying phase than in the short inundation phase (F = 4.83, p = 0.015; F = 5.95, p = 0.018) and in the long inundation (F = 6.60, p = 0.045; F = 5.92, p = 0.045) in the floodplain soil and river sediments, respectively.

Cohen's d was calculated to compare the effect sizes of the different phases on N₂O emissions derived from nitrification and denitrification (Supplementary Table 9). Overall, the effects of drying and short inundation were larger than the long inundation, whereby short inundation had stronger effects during the first cycle (N₂O emissions derived from nitrification and denitrification), and drying had greater effect size during the second cycle (N₂O emissions derived from denitrification).

The relative effect of drying and rewetting events on nitrification and denitrification was calculated as the gas flux change rates between long-term inundation and drying-rewetting treatments as well as between the drying phase and short-term inundation. During cycle 1, the short inundation increased N₂O emissions from denitrification at all sites. Flux changes were highest in river sediments (167%), followed by the parafluvial sediments (125%) and floodplain soil (13%). The highest flux change occurred for nitrification (529%) during the short inundation in the parafluvial sediments. During cycle 2, increases in N₂O emissions from both nitrification and denitrification were observed at all sites during the drying phase. Flux increases were highest in the floodplain soil (387 and 350%, respectively), followed by the river (180 and 271%, respectively), and parafluvial (38 and 155%, respectively), sediments.

At the sites not frequently exposed to drying-rewetting cycles, i.e., the floodplain soil and river sediments, the fluxes from both nitrification and denitrification were more affected by drying than in the parafluvial site. Flux changes were highest in the floodplain soil (1,947 and 1,100%), followed by the river sediments (720 and 1,013%) and the parafluvial sediments (50 and 211%), for nitrification and denitrification, respectively. During cycle 1, the parafluvial sediments were more impacted by the short inundation event (175 and 260%) than the floodplain soil (−6% and 125%) and the river sediments (−46 and 60%) for nitrification and denitrification, respectively. During cycle 2, the N₂O fluxes from nitrification and denitrification decreased at all sites during the short inundation event.

The forward selection procedure indicated that: OM content, DOM availability and quality (DOC, component 3 and FI), the frequency (C1 and C2) and phase (drying and inundation), were the key environmental factors relating N₂O fluxes (Figure 7; Supplementary Table 10). Approximately 42% of the flux variability could be explained by the selected environmental variables (constrained variance: 42%, unconstrained variance: 58%; F = 10.8; p = 0.001), with the first axis explaining a significant part of the variance of the response variables (F = 75.6; p = 0.002). The angles between all vectors on the RDA triplot reflect their linear correlation. Namely, the higher fluxes observed during the drying phase (cycle 2) were related to changes in DOM quality (with respect to denitrification) and to higher DOC and OM contents explaining higher fluxes derived from nitrification, after the first drying-rewetting cycle. Lower N₂O fluxes during the inundation phase were concurrent with lower FI values.

We hypothesized that drying-rewetting cycles in aquatic systems enhance N₂O from both nitrification and incomplete denitrification as these events affect key variables influencing N₂O cycling, namely substrate availability and DOM quality. Similar to other studies, NO3- tended to be higher during the drying phase and lower during the inundation, while NH4+ tended to be lower during drying and higher during wet conditions. These shifts are indicative of enhanced N processing rates. The increase in NO3- content and decrease of NH4+ most likely results from OM mineralization and continuous nitrification, throughout the drying period (16, 17, 19). Higher denitrification rates have also been detected between the onset of drying as a result of increased net nitrification rates, as well as N mineralization rates (19, 51).

Nitrification was an important source of N₂O emissions. Studies have shown that nitrification rates can be equal or greater than denitrification rates in inland waters (3, 52, 53), and in saturated wetland soils and after drainage and often plays an underappreciated contributing to N₂O emissions under these conditions (27). We observed highly enhanced N₂O emissions from both nitrification and denitrification following drying events, although these peaks were only apparent after repeated drying during cycle 2. This pattern suggests that the transition from wet to dry phases has a high potential for N₂O pulses from nitrification and denitrification (Figure 6, cycle 2), especially if these cycles are recurring. This trend has been observed previously in other studies. In soils, the highest N₂O emission peak intensities have also been shown to occur during the drying phase compared to the wetting (near-saturation) phase and were correlated positively with the amount of water drained (54). Similarly, enhanced N₂O emissions were measured during the transition from flooded to dry sediment conditions in a seasonal floodplain lake (51). The high N processing activity during initial drying was attributed to high spatial heterogeneity in redox conditions. Koschorreck (51) assumed that the patchy distribution of oxic and anoxic microsites between flooding and sediment drying increased the oxic-anoxic boundary area where coupled nitrification-denitrification could occur, increasing N₂O emissions.

Drying-rewetting cycles drive environmental gradients, affecting N₂O production and emission, namely O₂ gradients, and substrate diffusion and availability and release. During these transitions, suboptimal conditions for both nitrification and denitrification can lead to an increase in N₂O emissions, for example, inhibition of N₂O reductase by O₂, which catalyzes the reduction of N₂O to N₂ during denitrification (55). Higher processing rates and N₂O fluxes are often observed when nitrification and denitrification processes are coupled [reviewed in (6)]. These fluxes are most likely driven by physical mechanisms related to the release of entrapped N₂O in an initial stage, and sustained by a period of enhanced microbial activity thereafter (10, 18, 56). Comparing the different sites, changes in flux rates due to drying point to N₂O pulses of a smaller magnitude in areas frequently exposed to drying-rewetting cycles, indicating that N₂O peaks may become less intense after several drying-rewetting cycles (11). These differences may relate to community compositions which are site specific (i.e., soil, long inundated sediments, temporarily inundated sediments), therefore different microbial associations may be favored at each site, and indicative of functional differences during environmental changes (57).

We acknowledge the presence of DNRA in the long inundation cores due to favorable conditions for this pathway, namely strictly anaerobic sediments and limited N-NO3- supply. However, we consider the influence from this pathway to N₂O emissions to be negligible compared to emissions from denitrification, as N₂O is not an intermediate product, but rather is only produced when N-NO2- is allowed to accumulate (13). In these conditions, the complete DNRA process with the end product of NH4+ might be more efficiency than the incomplete NO3- reduction that produces N₂O (42), which is in agreement with the increased N-NH4+ concentrations (Figure 3).

Drying-rewetting cycles affect C content and quality (e.g., 20) in addition to N availability and N cycling pathways (11). The variation in N₂O fluxes was best explained by OM content and availability (nitrification) and DOM quality (denitrification). DOM is a complex mixture of organic compounds and is typically the most abundant source of C (DOC) in aquatic systems (58), fueling microbial processes, including denitrification (59). It results from a combination of allochthonous and autochthonous sources and their microbial processing (60). In rivers, most DOM is derived from allochthonous sources (soil and plant material) (58). Here, we observed that DOM is more terrestrially derived (FI ~ 1.2), not only in soil but also in parafluvial and river sediments, although the DOM composition differed between soils and sediments.

Results from the DOM quality support the hypothesis that hydrologically driven changes in DOM favor N₂O emissions. In this study, the increased N₂O fluxes from denitrification during the second drying phase were related to DOM quality changes specific to the tryptophan-like peak. This correlation suggests a change in the DOM pool during drying-rewetting cycles. This link seems to originate from the quality of the DOM pool, and related to the dependence on readily available organic carbon, which influences C and N turnover processes. Because most denitrifiers are heterotrophic, the potential for N₂O production via denitrification increases, with more labile forms of DOC supporting higher denitrification rates [(13) and references therein; (61); accepted]. Hence the role of DOM depends not only on the quantity but also on its quality. The variability in hydrological connectivity plays an essential role in DOM composition (22, 23), with protein-like components achieving maximum abundances in the littoral zones (62). Therefore, in hydrologically dynamic areas, the high availability of protein-like substrates suggests a high DOM turnover by heterotrophic microbes. Indeed, a positive correlation between the presence of protein-like fluorophores and denitrification rates and N₂O fluxes has been previously observed in fluvial ecosystems (14, 24).

Our study observed an increase in fluorescence intensity for a protein-like component during transitions from short inundation to the drying phase (Figure 5). Component 3 shows protein-like characteristics similar to freshly produced tryptophan, consistent with others previously reported and well-matched with the Open Fluor database with similarity scores (Tucker's congruency coefficients) of > 0.95. In particular, a similar component was identified in a study investigating the driving role of hydrological connectivity on DOM processing [(21), Supplementary Table 6]. The study identified periods with disproportionately high rates of DOM processing (hot moments), one of which was characterized by a short pulse of protein-like, autochthonous DOM net release at the beginning of hydrological disconnection. Similarly, other studies have found higher protein-like fluorescence as drought proceeds, indicating a shift to more tryptophan-like, non-humified, bioavailable DOM (22, 23).

DOC content was also generally higher during the drying than in the rewetting phase. Increases in DOC quantity and DOM bioavailability have been reported in streams during initial surface drying (21, 23, 63). Together with an increase in DOC concentration, these changes in DOM composition are likely related to microbial cell lysis and/or DOM exudation in response to stress conditions (10, 23). Therefore, these phases could be considered biogeochemical hot moments (4). Lower FI fluorescence observed during the inundation phase may indicate that a greater relevance of allochthonous sources exists, most likely originating from soil and sediment leachates (22). Lower N₂O emissions were likely the combined result of complete reduction of N₂O to N₂ by denitrification and exhaustion of NO₃ and/or NO₂. Overall, evidence suggests that dissolved organic matter processing, composition (i.e., aromaticity), and quality are affected by drying-rewetting cycles. These cycles are linked to higher N₂O emissions, with fewer aromatic compounds contributing to OM mineralization and N₂O pulses.

Nitrification was not related to NH4+ concentrations but to soil and sediment OM content and availability (DOC) during drying. These results suggest that NH4+ was provided to nitrification via N mineralization (52), increasing N₂O emissions through this process. Although nitrification is a chemo-autotrophic process, organic carbon can be an important regulator of nitrification rates via environmental C/N ratios and competition between heterotrophic and nitrifying bacteria (64, 65). Indeed, DOC availability decreased between drying cycles and the higher N₂O fluxes observed for nitrification during the second drying phase could be related to a less inhibitory effect of organic carbon.