Experimental flooding shifts carbon, nitrogen, and phosphorus pool distribution and microbial activity

Flooding transforms the soil environment, impacting small-scale controls on mineral associated carbon (C), nitrogen (N) and phosphorus (P) persistence and mobilization. Yet during flood events, mineral associated C, N, and P may not respond in the same way, such that soluble C, N, and P concentrations and stoichiometry may change potentially impacting microbial activity. Using a laboratory incubation approach, we investigated how flooding impacts C, N, and P pool distribution and microbial activity across a 1-week flood event and after drying. We found that all three mineral associated pools responded dynamically to flooding, increasing and decreasing throughout the flood with a 5.9% increase in mineral associated C and 32.5% decrease in residual P post flood. However, mineral associated C, N, and P each shifted at different temporal points, indicating that they are likely responding to separate destabilization mechanisms working at different temporal scales. Soluble C increased by 57% and soluble N decreased by 72% at the beginning of the flood which remained post-flood. However, soluble P behaved more similarly to the mineral associated pool, shifting throughout the flood period. The microbial community maintained and even increased their exo-cellular activity throughout the flood period. Our research demonstrates that the mineral associated pool can shift with short-term flooding altering the composition and quantity of the soluble pool and microbial activity.


Introduction
With climate change altering precipitation patterns, the extent and duration of floods are expected to increase in many parts of the world, with more land flooding annually and remaining flooded for longer (Hirabayashi et al. 2013;Jeong et al. 2013). Flood events saturate the soil which drastically changes the environment and may shift small-scale controls on mineral associated carbon (C), nitrogen (N), and phosphorus (P) persistence (Huang and Hall 2017;Lamparter et al. 2009). Here we consider soil flooded when it is fully saturated with standing water on top. The mineral associated pool refers to inorganic and organic matter that is associated with silt and clay sized particles largely consisting of small bipolymers and monomers (Lehmann and Kleber 2015;Sokol et al. 2022). This pool is important for C, N, and P retention with turnover times from decades to centuries and is thought to be largely biologically unavailable (Hemingway et al. 2019;Miller et al. 2001). Yet evidence suggests that the mineral associated pool is more dynamic than previously thought, showing rapid changes due to environmental disturbances like increases in moisture or flooding (Huang and Hall 2017). Such flood events can change how inorganic and organic matter is partitioned between the mineral associated and soluble pools. C, N, and P may be impacted differently by flooding depending on how they are protected within the soil and the rate of microbial C, N, and P uptake. Soil C, N, and P cycling are often linked, where a change in one pool influences the behavior of the others. However, their potential varying responses to flooding are not well explained as they are typically examined separately (Mooshammer et al. 2014).
The uncertainty in how flooding affects the simultaneous and often interdependent responses of C, N, and P pools limits how we predict and understand patterns of soil processes and ecological functioning in response to flooding. For instance, flood-induced changes to the mineral associated C, N, and P pools could change element retention, distribution, and stoichiometry in the soil with subsequent impacts on plant and microbial community, C and nutrient availability (Mooshammer et al. 2014;Mori et al. 2018;Tian et al. 2017).
Mineral associated C, N, and P can be mobilized by flooding through three key mechanisms: changes in redox potential, disintegration of aggregates, and increases in biological accessibility. With flooding, soil becomes temporarily waterlogged, often lowering the redox potential (Chen et al. 2019;Shaheen et al. 2021). Soil C, N, and P are stabilized within their respective mineral bound pools by bonding with clay minerals and Fe and Al oxy(hydro)oxides (Baldock and Skjemstad 2000;Hemingway et al. 2019;Kleber et al. 2015). Low redox conditions can result in reductive dissolution of mineral surfaces where, for instance, Fe oxides are reduced, causing mineral associated C, N, and P to desorb into their respective soluble pool (Bailey et al. 2019;Lin et al. 2018). In unsaturated conditions smaller, less connected intraaggregate micropores protect more complex forms of C and nutrients through physical isolation and anoxic microsites (Cates et al. 2022;Keiluweit et al. 2017). Flooding can break up these aggregates, releasing physically protected organic matter into the soluble pool (Fierer and Schimel 2003). Once in solution, C, N, and P are more mobile in the soil pore water which can transport previously protected compounds into groundwater or to biologically accessible areas of the soil where C and N can be lost through respiration (Bailey et al. 2019;Marschner and Kalbitz 2003).
C, N and P may respond differently to these mobilization mechanisms due to how they are primarily stabilized in the soil. Soil C and N are predominantly protected through biotic processes, while P is stabilized principally through abiotic processes (Celi et al. 2022). Bio-inaccessibility plays a central role in C and N persistence. For example, the centers of microaggregates, low moisture conditions limiting pore space connectivity, and anaerobic microsites each limit microbial access and/or ability to process organic matter and thus are critical for C and N retention (Bailey et al. 2019;Celi et al. 2022;Keiluweit et al. 2017). In contrast, sorption/desorption dynamics are essential for P retention. Inorganic P and phosphorylated organic compounds form organo-mineral interactions rapidly and are bound tightly (Guppy et al. 2005;Ruttenberg and Sulak 2011). On the other hand, C, and to some degree N, are more loosely held than P on the soil matrix and thus can be destabilized via microbial attack (Jilling et al. 2018). Thus, C and N are more likely to be impacted by changes in pore space connectivity and microbial activity, while P is more likely to be impacted by shifts in redox conditions during flooding.
Following an increase in the concentration and mobility of soluble and biologically available substrates under saturated conditions, microbial uptake of C, N, or P may increase, with the relative C, N, and P assimilation depending on microbial stoichiometric demands (Buchkowski et al. 2015;Mooshammer et al. 2014). However, more prolonged flooding and the resulting increases in anoxia may limit microbial exo-cellular enzyme production and assimilation of newly biologically available substrates as anaerobic conditions limit the thermodynamic energy yield of decomposition (Dyckmans et al. 2006;Torbert and Wood 1992;Unger et al. 2009). During short flood periods when there are still aerobic microsites, microbes may be able to benefit from both the increases in substrate availability and aerobic conditions (Maranguit et al. 2017).
Differences in flood duration may also affect the magnitude of desorption. During short floods or early flood onset, spatial heterogeneity exists where aerobic and anaerobic soil microsites co-occur (Dorau et al. 2022;Schlüter et al. 2022). Increases in flood duration allows more time for water to percolate, causing greater and more uniform waterlogging throughout the soil matrix (Reddy and DeLaune 2008). As such, longer flood periods should create a more reduced and anoxic soil environment, leading to more desorption of mineral associated C, N, and P. Thus, with longer flooding, higher concentrations of soluble compounds may be a result of both greater desorption as well as inhibition of microbial aerobic respiration and uptake of substrates.
When the soil dries following flood retreat and returns to pre-flood conditions soluble C, N, and P can re-enter the mineral associated pool. However, the disturbance from flooding and drying may affect the C, N, and P mineral pool composition due to changes in both the chemistry and stoichiometry of the soluble pool and resorption dynamics (Guggenberger and Kaiser 2003). For instance, as soil matric potential declines, N-and P-rich cytoplasmic material can be released following cell lysis and the microbial production of N-rich osmoregulatory solutes (Fierer and Schimel 2003;Warren 2016). Soluble C, N and P concentrations post-flood may also be affected by changes in microbial exo-cellular enzyme activity and C, N, and P microbial assimilation during flooding (Gu et al. 2019). These releases and uptake change the C:N:P of compounds available for resorption to the mineral associated pool. Competition for sorption sites of oxidized mineral surfaces, may further shift the C:N:P ratio of the mineral associated pool compared to pre-flood conditions (Bird et al. 2008;Guppy et al. 2005). Both N and P are thought to outcompete C for sorption sites, driving the outcome of resorption (Hatton et al. 2012;Schneider et al. 2010;Sollins et al. 2006).
While flood events are expected to influence the partitioning of organic matter and nutrients between the mineral associated and bioavailable soluble pool; it is difficult to predict the impacts of flooding on C, N, and P given the inter-relatedness between pools and among the three elements. Thus, it is necessary to examine how flooding and its duration impact C, N, and P both independently and concurrently throughout a flood period and post-flood to understand the cascading effects of flooding on the soil system. To better characterize and explain flood effects on soil C and nutrients, we examined the partitioning of C, N, and P across soil pools and microbial biomass as well as exo-cellular activity responses during a weeklong flood and after drying in a soil incubation. We hypothesized that as flood duration increases, mineral associated C, N, and P and microbial activity will decrease and in turn soluble C, N, and P will increase. However, post-flood microbial activity will increase relative to the flood period, with longer flood periods showing a greater post-flood microbial response due to higher bioavailable, soluble C, N, and P. Lastly, we expect hysteresis with drying because of changes in the concentration and composition of the soluble and mineral associated pools during the flood, and thus a net change in post-flood mineral associated C, N and P composition compared to pre-flood conditions.

Incubation design
We collected soil from the Emile A. Lods Agronomy Research Centre in Sainte-Anne-de-Bellevue, Québec (latitude: 45°25′N; longitude: 73°55′W; 39 m elevation) in October 2019 approximately 2 weeks prior to corn harvest at a depth of 0-15 cm. The field was planted with corn (Zea mays) at the time of sampling and is under a corn-soybean (Glycine max L merr.) rotation using standard agronomic practices of the region. The site has a mean annual temperature of 6.8 °C and a mean annual precipitation of 1000 mm (Environment Canada 2019). The soil is of the Châteauguay series with 33% clay content, a pH of 6.7 and a field water holding capacity of 35%. Mean concentrations of total iron (Fe), aluminum (Al), and calcium (Ca) were 225 ppm Fe, 919 ppm Al, and 1848 ppm Ca. The soil was sieved to 2 mm at field moisture conditions and visible roots were removed before the incubation period.
Our experimental design consisted of 3 sampling periods pre-flood (control), during flood (flooded soil) and post-flood (dried back to pre-flood waterfilled pore space). For the during flood period, we destructively sampled three times, at 0.5 hour, 24 hours, and 1 week of flooding. Each of the 3 flood sampling time points also had a corresponding postflood sample set that was sampled after drying to pre-flood moisture conditions. Each sampling time point was replicated 5 times, except for our control (pre-flood moisture conditions) which was replicated 10 times (n = 40, Fig. 1). Pre-incubation field fresh soil was at 23% water-filled pore space, equivalent to 20% gravimetric water content. We placed approximately 100 g (dry weight) of the homogenized fieldfresh soil in cylindrical columns (10 cm length and 4 cm diameter) with fine mesh (< 0.53 µm) attached to the bottom to allow water to drain and parafilm on top to allow oxygen exchange while preventing water loss. The columns stood inside cups on a flat surface to collect leachate and allow for resorption. To represent waterlogged soils characteristic of seasonal flooding, soils were brought up to 100% water-filled pore space and then inundated under water during the incubation with deionized water. Soil was wetted both from below and above to simulate the simultaneous increase in moisture from ground water rise and soil surface ponding that soils often experience during seasonal flooding. Flooded conditions were maintained by adding water from above twice a day or as needed. We subsampled soils for C, N, and P, and microbial analyses immediately after each flood sampling time point. Another subset of soil incubations was dried to 23% water-filled pore space and then sampled for the same analyses for our post-flood treatment. During the incubation period, soil columns were kept between 18-20 °C in the dark. Throughout the incubation and drying period, leachate was collected to represent cumulative leachate. Leachate and all soils after sampling were kept frozen at − 20 °C until analyses.

Soil carbon and nitrogen pools
We measured total, mineral associated, and water extracted soil organic C and total N in all soil samples. To obtain water extracted organic C (WEOC) and total N, we added 40 ml of deionized water to 10 g (fresh weight) slowly thawed soil, shook for 30 minutes on an end-to-end shaker, centrifuged for 20 minutes at 3600 g and then filtered to 2.5 µm. Here we operationally define the water extracted C and N at < 2.5 µm as our soluble pools. While a water extracted pool will include non-soluble compounds, it better captures all organic matter that might be more mobile in water and thus more biologically available. Water extracted total organic C and total N concentrations were then determined on a TOC/N analyzer (Shimazdu Corp, Kyoto, Japan). Water-extracted nitrate and ammonium were determined on water Fig. 1 Incubation sampling scheme. Each soil column in the figure represents 5 replicates, except for pre-flood conditions which represents 10 replicates (n = 40). Replicates were destructively sampled at the end of the designated flood or drying period extracts by spectrophotometric analysis (Doane and Horwáth 2003;Hood-Nowotny et al. 2010), using a 96-well spectrometer at 540 nm for nitrate and 650 nm for ammonium (BioTek Instruments, Winooski, Vt, USA). Water extracted organic N (WEON) was determined by the difference between total and inorganic N.
We extracted mineral associated organic C (MAOC) and total N (MAON) by size fractionation (Cambardella and Elliott 1992;Jilling et al. 2020;Sokol and Bradford 2019). Soil samples were dried at 105 °C for 48 hours and then dispersed by shaking 8 g dry soil in 40 ml of 0.5% sodium hexametaphosphate on an orbital shaker for ~ 16 hours. We isolated the clay and fine silt fraction (< 53 µm) by rinsing the dispersed soil with deionized water over a 53 µm sieve. The < 53 µm fraction, representing the MAOC and MAON pool, was then dried for 48 hours at 105 °C and finely ground by mortar and pestle. We quantified the total organic C and N of the isolated mineral fraction using a flash combustion elemental analyzer (Costech EA ECS 4010, Valencia Ca, USA). Ovendried (105 °C) and finely ground bulk soil was also analyzed for total soil organic C and total soil N. Inorganic C in our soil was considered a negligible fraction of total C as determined by the fizz test (Baldock et al. 2013).

Soil phosphorus fractionation
We used a modified Hedley sequential fractionation to analyze nine operationally defined P pools, representing differences in chemical stability (Tiessen and Moir 1993). Samples were extracted stepwise in a sequence from most to least available: first with deionized water, then 0.5M NaHCO 3 (pH 8.5, targeting moderately available P forms), 0.1M NaOH (targeting Fe and Al bound P), and 1M HCl (targeting Ca bound P). Residual P was considered the fraction remaining in the soil after the final extraction and was determined by hot acid digestion. We consider the water extracted pool to be the soluble pool and the total NaOH extracted, inorganic HCl extracted and residual pools to be the mineral associated pools. Briefly, 0.5 g (fresh weight) of slowly thawed soil was extracted with 30 ml of extractant solution, shaken on an orbital shaker overnight (~ 16 hours), centrifuged at 3600 g for 20 minutes and decanted. This procedure was repeated with each extractant. To obtain total P in the fractions within the 0.5M NaHCO 3 and 0.1M NaOH extracts and the residual P pool, we performed a hot acid digestion. The extracts and soils were digested with concentrated H 2 SO 4 and 30% H 2 O 2 at 360 °C for 2 hours. All extracts were analyzed on a 96-well plate spectrometer (BioTek Instruments, Winooski, Vt, USA) at 630 nm using malachite green to determine P concentrations (Ohno and Zibilske 1991). We obtained total soil P through the addition of all soil P pools.

Exo-cellular enzyme activity and microbial biomass
We determined potential exo-cellular enzyme activity in the different flood and post-flood treatments to characterize microbial activity and responses to changes in substrate availability (Saiya-Cork et al. 2002). We targeted four exo-cellular hydrolytic enzymes that act on bioavailable C, N, and P substrates: ß -1,4-glucosidase (BG; targeting the final step of microbial cellulose breakdown), leucine and tyrosine amino peptidases (LAP, TAP; targeting N compounds associated with amino acids) and acid phosphatase (PHOS; targeting organic P). Total peptidase activity is reported as a sum of LAP and TAP. These hydrolytic enzymes were measured in soil slurries consisting of 1 g fresh weight slowly thawed soil and 50 mM sodium acetate buffer homogenized by blending. The buffer was adjusted to the average pH (6.5) of all soil incubations. Aliquots of the soil slurry, fluorescing substrates, and methylumbelliferyl or methylcoumarin standard were added to a black 96 well microplate. Potential enzyme activity was measured fluorometrically on a fluorometer, excitation fluorescence was set to 365nm, and emission intensity was set to 450 nm (BioTek Instruments, Winooski, Vt, USA) and results are presented as activity (nmol h −1 g −1 ).
We measured microbial biomass C (MBC) and N (MBN) using a modified chloroform fumigation extraction method (Wu et al. 1990). Briefly, 8 grams of slowly thawed soil was fumigated with chloroform for 24 hours in the dark and then extracted with 40 ml of 0.5M K 2 SO 4 . A corresponding set of samples was extracted in the same manner but not fumigated. Fumigated and unfumigated extracts were analyzed on a TOC/N analyzer (Shimazdu Corp, Kyoto, Japan). MBC and MBN were calculated as the difference between the fumigated and unfumigated C and N concentrations, these values were further adjusted with a K EC (extraction efficiency) of 0.45 for C and 0.54 for N. Soil samples were also analyzed for microbial biomass P using chloroform fumigation followed by 0.5M NaHCO3 extraction (Brookes et al. 1982). However, due to methodological issues that generated unreliable results, data are not presented.

Data analysis
Statistical analysis was performed in R v 4.1.1 (2021-08-10) using the dplyr, vegan and glme packages for data processing and analyses and ggplot2 for data visualization. For all statistical analyses, significance was defined as p < 0.05. We determined differences in flood duration within an individual variable using one-way ANOVAs followed by a Tukey post hoc test. Correlations between variables were analyzed using generalized linear models. Normality and homoscedasticity of residuals was evaluated with diagnostic plots and a Shapiro-Wilk test. Data were log or square root transformed, and outliers removed as necessary.
Principal coordinate analysis (PCoA) was performed on the mineral associated pools (total 0.1M NaOH extracted, inorganic 1M HCl extracted and residual for P), water extracted pools and microbial biomass C and N for 1) during flood treatments and 2) post-flood treatments. Distances were calculated using the Bray Curtis index in PC-ORD version 7.09 (McCune and Medford 2016). Vectors are shown for variables with R 2 greater than 0.3. The influence of flood duration on these ordinations was tested using Permanova and correlation among variables was determined with Pearson correlation analysis ( Fig. S1-2).

Results
Mineral associated C, N, and P response to flooding across a 1-week flood period Flooding induced shifts in mineral associated C, N and P pools (Fig. 2). However, both increases and decreases were observed at different time points during the flood for each element, indicating that mineral associated C, N, and P are responding distinctly to flooding. There were also differences in the variability between replicates among the mineral associated pools, with the more dynamic pools showing larger variability (e.g. NaOH-P verse HCl-P).
Flooding decreased the amount of MAON from 1,215 mg kg −1 pre-flood to 1,056 mg kg −1 after 1 week ( Fig. 2b;  We compared three different mineral associated P pools─total NaOH extracted (NaOH-P), inorganic HCl extracted P (HCl-P), and total residual P─that represent the least soluble and most stable forms of P in our Hedley fractionation (Fig. 2c-e). There was a strong response to flooding in the NaOH-P (p = 0.006) and residual P fractions (p = 0.008), exhibiting variability over the course of the flood and following drying. For instance, after 0.5 h of flooding, NaOH-P increased from 409 mg kg −1 preflood to 573 mg kg −1 , followed by a decrease to 423 mg kg −1 after 24 h and then an increase to 600 mg kg −1 after 1 week. The NaOH-P pool also decreased with drying post-flood compared to the 1-week flood time point (p = 0.027) but returned to pre-flood concentrations. Residual P was the only P pool where we observed a net change in P over the flood event, with a decrease from 186 mg kg −1 pre-flood to 125 mg kg −1 (p = 0.040) and finally to 200 mg kg −1 at the 1-week time point (p = 0.025). The HCl-P pool did not respond to flooding (p > 0.05).
Water extracted C, N, and P response to flooding across a 1-week flood period We measured the WEOC, WEON, and water extracted inorganic P (WEP) pools to understand how flooding and the subsequent shifts in the mineral associated pools impact the corresponding soluble pools (Fig. 3). The WEOC pool sharply increased and the WEON pool decreased at the beginning of the flood period and these changes remained throughout and post-flood. WEOC increased by 57% from 20 to 32 mg kg −1 after 0.5 h, increased again to 36 mg kg −1 after 1 week and remained at 34 mg kg −1 postflood, an overall increase of 66% ( Fig. 3a; p < 0.001). In contrast, WEON decreased from 20 to 5.6 mg kg −1 with 0.5 h of flooding, and this 72% decrease remained post-flood ( Fig. 3b; p < 0.001).
The WEP pool showed a more dynamic response to flooding compared to WEOC and WEON, shifting throughout the flood period like the mineral associated P pools, especially NaOH-P ( Fig. 4c; p < 0.001). It is worth noting that WEP also showed a higher level of variability between replicates compared to the other soluble pools, especially WEON. Flooding caused WEP to increase with 0.5 h and 1 week of flooding compared to both pre-flood and 24 Fig. 2 Mineral associated organic C (a), mineral associated N (b), total 0.1M NaOH extracted P (c), inorganic 1M HCl extracted P (d) and total residual P (e) concentrations (mg kg −1 dry soil) for pre-flood, 0.5 hour, 24 hours, and 1 week of flood-ing and post-flood. Boxplots show median, maximum, minimum, and outliers. Y-axis scales differ between panels. Different letters indicate significant pairwise differences (p < 0.05); n = 5 except for pre-flood where n = 10, d.f.=4 h of flooding. Post-flood the WEP pool remained at a higher concentration compared to pre-flood conditions, however this difference was not significant such that there was no overall net change in WEP with the flood event.

Flooding impacts on the microbial activity and biomass
We observed that microbial exo-cellular enzyme activity is maintained and, in some instances,  Fig. 4 Exo-cellular enzyme activity (nmol h −1 g −1 dry soil) ß -1,4-glucosidase (a), sum of leucine and tyrosine amino peptidases (b) and acid phosphatase (c) at pre-flood, 0.5 hour, 24 hours, and 1 week of flooding, and post-flood. Boxplots show median, maximum, minimum and outliers. Y-axis scales differ between panels. Different letters signify significant differences (p < 0.05); n = 5 except for pre-flood where n = 10, d.f.=4 increases in response to flooding and the subsequent changes in substrate availability (Fig. 4). The BG enzyme activity paralleled the soluble C pool, increasing from pre-flood 124 to 155 nmol h −1 g −1 at 0.5 h which remained at a higher level throughout the flood period ( Fig. 4a; p = 0.002). However, post-flood, BG activity returned to pre-flood levels. Peptidase (sum of LAP and TAP; Fig. 4b) activity fluctuated throughout the flood period returning to pre-flood activity with drying post-flood. PHOS activity gradually increased during flooding though not significantly during the first 24 hours. After 1 week, PHOS activity increased from 311 nmol h −1 g −1 pre-flood to 374 nmol h −1 g −1 (Fig. 4c; p = 0.004). Following drying, we observed a decrease in PHOS activity to 255 nmol h −1 g −1 compared to pre-flood conditions (p = 0.016) and 1-week conditions (374 nmol h −1 g −1 ; p < 0.001).
We found that changes in the water extracted, and potentially bioavailable substrates correlated with exocellular enzyme activity, with WEP correlating to all three enzyme groups (C-, N-, and P-cycling; Table S1). BG activity was most impacted by changes in the soluble pool, correlating positively with WEOC and WEP and negatively with WEON. In contrast, both peptidase and PHOS activity were only positively correlated with WEP. However, the low r 2 values (0.17) for the correlation between PHOS and WEP suggest that changes in P availability explain little of the observed variation in PHOS activity.
Flooding had limited effects on microbial biomass C and N. Flooding did not change MBC or MBN during or after the 1-week flood period (Table S2; p > 0.05). However, microbial biomass C:N decreased to 5.3 after drying compared to pre-flood (16.6) and 1 week of flooding (21.6), driven by both a decrease in MBC and increase in MBN (Table S2; p = 0.005).

Impacts of flood duration on post-flood C, N and P stocks
We analyzed how flood duration impacts the effects of drying on C, N, and P dynamics by examining the percent change between pre-and post-flood conditions (Fig. 5) for each flood duration in the mineral associated pools, water extracted pools, and exo-cellular enzyme activity. Based on post-hoc tests, flood duration impacted the magnitude and direction of drying induced C, N, and P shifts.
Drying differentially impacted the mineral associated C, N, and P pools causing the greatest change in the mineral associated P pools. Drying increased MAOC only under the longest flood duration compared to pre-flood concentrations (p < 0.001; 5.9%). MAON did not shift with drying under any investigated flood scenario. NaOH-P decreased with drying compared to pre-flood conditions only with the 0.5 h flood duration (p = 0.002; 42.6%). In contrast, residual P decreased with drying compared to pre-flood conditions only under the longest flood duration (p = 0.039;

Fig. 5
Percent change between pre-to post-flood for 0.5 hour, 24 hours and 1 week flood durations for mineral associated pools, soluble pools and exocellular enzyme activityß-1,4-glucosidase (BG), leucine and tyrosine amino peptidases (peptidase) and acid phosphatase (PHOS). Asterisks represent significant differences determined with an ANOVA followed by a Tukey post hoc test (p < 0.05), n = 5 except for pre-flood where n = 10, d.f.=6 32.5%). The HCl-P pool showed no change after drying for any flood scenario. When compared to preflood soil, WEOC increased post-flood while WEON decreased under all flood scenarios. WEP only increased with drying compared to pre-flood conditions in the 0.5 h flood period (p = 0.013; 34.9%). BG and peptidase activity did not change with drying compared to pre-flood conditions under any flood scenario. However, PHOS activity decreased after 1 week of flooding compared pre-flood levels. We did not include microbial biomass in this analysis as it did not shift with drying under any flood scenario.

Soil pool composition during and after flooding
To understand the relationship among flood-induced effects in the C, N, and P pools and the variables driving these compositional shifts we conducted a PCoA ordination with the mineral associated pools, water extracted pools, MBC, and MBN for both during and post-flooding. During flooding, each flood duration had distinct C, N, and P pool composition ( Fig. 6a; p = 0.002). Most of the separation occurred along PCoA 1 (74.9%) between the 0.5 h and 1 week flood periods primarily driven by WEOC and MAOC. MAOC was positively associated with PCoA 1 (tau = 0.89), corresponding to the 0.5 h flood period, while WEOC was negatively associated with PCoA 1 (tau = − 0.648), aligning with the 1-week flood duration. The 24 h flood period corresponded to PCoA 2 (13.8%), driven by primarily by MAON (tau = − 0.538), NaOH-P (tau = 0.692), and WEP (tau = 0.516). Three P pools, WEP, NaOH-P, and residual P, were positively associated with both the 0.5 h and 1 week flood treatments.
Post-flood, the 1-week flood duration treatment was distinct in pool composition compared to the 0.5 h and 24 h flood treatments ( Fig. 6b; p = 0.002). No differences were observed in the dry down effect between the flood durations of 0.5 h and 24 h (p > 0.05). Most of the separation between the 1-week flood duration and the two shorter flood periods occurred along PCoA 1 (90.8%), driven by MAOC (tau = − 0.886) and MBC (tau = 0.524), with the 1-week flood duration associating with higher MAOC. The 0.5 h flood duration also separated from the longer flood periods along PCoA 2 (4.4%), associated with higher MAON concentrations (tau = 0.276). The 24 h and 1-week flood durations were negatively associated with NaOH-P (tau= − 0.714) along this axis.

Discussion
We investigated how flooding shifts mineral associated C, N, and P, and the cascading consequences for the soluble pools and microbial exo-cellular enzyme activity across a 1-week flood and after drying. We expected that as the flood progressed mineral associated C, N, and P would decrease resulting in an increase in water extracted elements. However, we found that each element and pool responded differently and more dynamically than expected. Our findings agree with recent work that shows the mineral associated pool can exhibit short-term fluxes with changes in soil environments like shifts in root exudate chemistry (Jilling et al. 2021;Keiluweit et al. 2015;Li et al. 2021), addition of labile substrates like phosphate (Spohn et al. 2022;Spohn and Schleuss 2019) and increases in soil moisture (Huang and Hall 2017). Overall, the flood event induced a change in the mineral associated pool composition, but only for C (and an increase in C:N) and residual P. We further anticipated that microbes would only initially respond to increased substrate availability when there were likely still aerobic conditions in the soil matrix. Unexpectedly, microbial enzyme activity was not limited by flooding and did not respond strongly to changes in substrate availability.
Mineral associated C, N, and P respond dynamically and differently to flooding During the flood period MAOC, MAON, NaOH-P, and residual P increased and decreased, indicating that these elements were moving in and out of the mineral associated pool (Fig. 2). Mineral associated C, N, and P each shifted at different temporal points of the flood. We suspect that differences in the way that C, N, and P are initially protected within the soil and their sensitivity to different destabilization mechanisms likely contributed to the variable C, N, and P response to flooding.
MAOC and MAON behaved more similarly to each other during the flood compared to mineral associated P, reaching their lowest concentration at 1 week of flooding. The late flood declines in MAOC and MAON pools may be explained by the release of physically occluded compounds and shifts in redox conditions. While our soils were sieved to 2 mm before the incubation period, microaggregates, critical for the long term protection of C and N, remained intact (Bingham and Cotrufo 2016). Water can take longer to infiltrate the centers of microaggregates, leading to a delay in anoxic conditions and corresponding lower redox, slowing the destabilization of mineral associated elements (Cates et al. 2022;Horn and Smucker 2005). This destabilization process may become more important as flood duration increases and likely contributed to the observed decreases of mineral associated C and N at 1 week. Additionally, higher pore space connectivity after 1 week potentially increased microbial contact with mineral surface, increasing the potential for biologically driven destabilization (Schlüter et al. 2022).
Desorption/resorption dynamics throughout the flood period might explain the mineral associated P responses to flooding. NaOH-P is especially prone to desorption under low redox conditions as it represents P bound to Fe and Al oxides (Tiessen and Moir 1993), and Fe 3+ can be reduced to Fe 2+ during periods of anoxia (Lin et al. 2018;Shaheen et al. 2021). Further, the decrease in nitrate (Table S4) during the flood period presumably caused increases in microbial-mediated reductive dissolution of Fe and the subsequent release of mineral associated P (Parsons et al. 2017). Though soil redox potential and O 2 concentrations were not measured, our soils were likely never completely anoxic, causing spatial and temporal differences in aerobic and anaerobic microsites. Such fluctuating conditions would create microscale shifts in redox conditions resulting in repeated desorption and resorption and frequent increases and decreases like those we observed. Over time, cumulative resorption may favor P, as P is thought to outcompete C for sorption sites (Guppy et al. 2005;Schneider et al. 2010). We hypothesize that in our systems, P may have sorbed onto mineral surfaces previously occupied by C and/or N, explaining why NaOH-P reached its highest concentration after 1 week, when MAOC and MAON were at their lowest. The 1M HCl-P pool did not respond to flooding at all, likely because this pool is usually associated with P bound to Ca, which is not sensitive to changes in redox potential and was in high concentrations in our soil (Ann et al. 1999).
Residual P is considered the most stable P pool and the most resistant to shifting redox conditions (Tiessen and Moir 1993). However, we found that this pool increased and decreased throughout the flood period (Fig. 3e). Previous studies suggest that the residual pool may decrease as concentrations of less stable pools (e.g. NaOH-P) decline, replenishing more soluble pools (Guo et al. 2000;Von Sperber et al. 2017). While we did not observe a strong relationship between the residual pool and WEP, we did not measure every P pool and the residual pool may have compensated for an unaccounted pool. When considering all three mineral associated P pools together our results indicate that the propensity of P to shift with flooding depends on how it is initially stabilized in the soil, with P associated with Fe (NaOH-P) most sensitive to shifts with flooding.
Drying results in a net change of the mineral associated pool composition and concentrations While the mineral associated pool was quite dynamic throughout the weeklong flood period, after drying several pools largely returned to pre-flood conditions (e.g. MAON, NaOH-P). However, we observed a small net increase in MAOC and MAOC:MAON ( Fig. 3a; Table S3) and a decrease in residual P postflood (Fig. 3e). As there was no change in total soil C or P with flooding compared to pre-flood conditions (Table S4), we suspect these changes represent a shift in pool distribution, partly explained by resorption dynamics post-flood.
Previous studies show that N and P outcompete C for sorption sites and thus we would expect a lower C:N and C:P following drying (Sollins et al. 2006;Spohn and Schleuss 2019;Violante and Gianfreda 1993). Yet, our observed increase in MAOC and decrease in residual P post-flood relative to pre-flood and higher MAOC:MAON (from 10.1 to 11.5 pre to post flood) contradict this. Our data show a hysteresis, similar to previous sorption-desorption studies (Bai et al. 2017;Chen et al. 2009;Okajima et al. 1983). Thus, unlike during the flood period, it may be that post-flood, sorption strength alone does not determine resorption dynamics, suggesting that sorption dynamics function differently during and post-flood (Chen et al. 2022;Jalali et al. 2018;Wu et al. 2022). Further, the relatively higher WEOC and lower WEON concentrations may have allowed C to outcompete P for available sorption sites (Jalali et al. 2018;Spohn et al. 2022), causing a net increase in MAOC and decrease in residual P post-flood.
Changes in the mineral associated pools are not uniformly reflected in the water extracted pools We expected that a decrease in mineral associated C, N, or P would be reflected in an increase in their respective water extracted pools (Cates et al. 2022;Lavallee et al. 2020). This was only the case for C, where the observed decrease in MAOC during flooding coincided with an increase in WEOC. However, we also found that, unlike the mineral associated pools, WEOC responded immediately to flooding, suggesting another potential source of WEOC besides the desorption from MAOC pool, at least during the early flood period. While we were not able to measure particulate organic matter due to small yields in our soil, we would expect it to still be present in small quantities, contributing to WEOC following its depolymerization (Olayemi et al. 2022).
In contrast to C which responded as expected, WEP increased and decreased in tandem with NaOH-P. It remains unclear why WEP and NaOH-P shifted together, and these shifts may be the result of changes to unaccounted P pools. For N, both WEON and MAON decreased during flooding, but this is likely due to system-level N losses from leaching where high quantities of N were found in the leachate (Table S5) or from denitrification resulting in nitrous oxide and dinitrogen emissions (Bronson et al. 1997;Reddy et al. 1984).
Flood duration impacts which elements respond to drying and soil C, N, and P pool composition In contrast to our hypothesis, flood duration did not universally increase the magnitude of pool responses, rather the direction and magnitude of the response depended on the element. Only WEOC and WEON showed similar changes at all flood durations with drying. Both WEP and NaOH-P showed net changes after drying with 0.5 h of flooding but not at longer flood periods, indicating that P is more sensitive to short flood periods. In contrast, MAOC only showed a net change after 1 week of flooding and MAON showed no changes after drying under any investigated flood scenario. These results further support our conclusion that C and N are destabilized through differing mechanisms compared to P.
We also found that flood duration greatly impacted C, N, and P pool composition during the flood (Fig. 6a). Flood duration had less of an effect on the C, N, and P soil composition with drying compared to during the flood, where only the weeklong flood differed from the two shorter time periods (Fig. 6b). Nonetheless, our findings demonstrate that relatively small differences in flood duration result in soil C, N, and P compositional changes.
Exo-cellular enzyme activity responds to changes in soluble C:N:P We anticipated that increases in water extracted compounds with flooding would prompt an increase in microbial activity and substrate assimilation but that the community would eventually be limited by lower O 2 . We found that microbes were able to maintain and even increase their enzyme activity throughout the flood. However, both MBC and MBN did not respond to flooding such that additional limitations to growth may have occurred.
With changes in the C:N:P of the bioavailable pool like those we observed, microbes need to adjust their exo-cellular enzyme production to maintain their stoichiometry when soil resources are imbalanced (Mooshammer et al. 2014). However, we did not find a relationship between soluble C:N:P and enzyme C:N:P (data not shown). Instead, it appeared that enzyme activity was primarily driven by P limitations. For instance, changes in WEP had the greatest impact on enzyme activity, positively correlating to the activity of all three enzymes (Table S1). In contrast, increases in WEOC were only associated with increased BG and not with peptidase or PHOS activity. Further, the large decreases in both inorganic and organic soluble N pools did not correlate to any change in peptidase activity. These results contradict previous studies that show that increases in available P decrease PHOS activity (Groffman and Fisk 2011;Margalef et al. 2021;Turner and Joseph Wright 2014), have no impact on BG activity (Turner and Joseph Wright 2014) and that soil N content is the main driver of PHOS activity (Margalef et al. 2017). Moreover, while drying largely returned exo-cellular enzyme activity back to pre-flood levels, PHOS was the exception where drying resulted in ~ 23% lower activity compared to pre-flood. The rapid resorption and desorption of P we suspect is occurring in our system may also be contributing to both P limitations and enzyme efficiency. Our results indicate that P remained the limiting substrate for microbial activity throughout the flood period despite increases in soluble P at 0.5 h and 1 week.

Ecosystem context and conclusions
We found that both water extracted and mineral associated C, N, and P as well as microbial activity shift with flooding and that the duration of the flood event impacts their response. Water extracted and mineral associated C, N, and P each responded differently in terms of magnitude, direction, and timing to flooding. These findings suggest that each pool and element are responding to different destabilization mechanisms acting at separate times during the flood and drying periods. Particularly interesting is the response of the mineral associated pools. While previous research finds that this pool is resistant to change, our findings agree with a growing body of evidence that the mineral associated pool responds dynamically and quickly to disturbance (Huang and Hall 2017;Jilling et al. 2020;Keiluweit et al. 2015;Li et al. 2021;Spohn et al. 2022).
Our findings also highlight that the response of mineral associated C, N, and P to flooding may be a consequence of their initial stabilization mechanisms. These stabilization mechanisms are dependent on both the behavior of the individual element and soil characteristics. Soil structure dictates both substrate and microbial location as well as the movement of O 2 and water (Hartmann and Six 2022). Our soils were collected from agricultural fields that undergo regular tillage practices, homogenizing the soil and disrupting macroaggregates (Young and Ritz 2000). Disturbances like this likely impact both the redox potential and the rate and degree at which water saturation occurs. Further, agricultural soils, including ours, are characteristically low in C and high in N compared to other temperate ecosystems, and typically a higher proportion of C, N, and P is mineral associated rather than physically protected compared to forests (Sokol et al. 2022). Thus, soils with different elemental concentrations or dominant protection mechanisms may exhibit flood responses different from what we observe here. Additionally, the ratio of Fe:Al:Ca is also an important predictor of the degree to which changes in redox, like those induced by flooding, impact mineral associated C, N and P (Celi et al. 2022). Soils with a higher Fe:Ca compared to ours may see greater shifts in the mineral associated pools then we did.
With more frequent flood events associated with climate change, our results suggest that different flood scenarios need to be considered as a potential control on the persistence of mineral associated pools as well as microbial C and nutrient availability. As soils act as an important C sink and there is extensive research on increasing the stable C pool to mitigate for climate change (eg. Bossio et al. 2020;Minasny et al. 2017;Paustian et al. 2016), determining how flooding may shift and even undermine these efforts is critical.
Acknowledgments We are grateful for the financial support for this research provided by the Natural Sciences and Engineering Council of Canada (NSERC)-CREATE Climate-Smart Soils grant (#528274-2019) funding HPL and MR and the Canadian Agri-Food Policy Institute 2020-2022 Doctoral Fellowship program to HPL. This work was also supported by a NSERC Discovery Grant (RGPIN-2021-03250) and a Fonds de recherche du Québec-Nature et technologies grant 2022-NC-297557 to CMK.
Author contributions HPL, CVS, and CMK conceived of and designed the study and HPL and MR carried out the experiment and collected and analyzed the experimental data. HPL wrote the first draft of the manuscript with contributions from CVS and CMK. All authors contributed to the interpretation of the data, writing and revising manuscript drafts and are accountable for the accuracy and integrity of all aspects of the work.

Funding
The authors have not disclosed any funding.

Data availability
The data and code that support the findings of this study are available on request from the corresponding author.