Soil texture and pH affect soil CO2 efflux in hardwood floodplain forests of the lower middle Elbe River

Floodplain ecosystems play a significant role in the global carbon (C) cycle, particularly due to their C sink potential in hardwood floodplain forests. However, in these forests, interactions between a heterogeneous micro‐relief and anthropogenic landscape changes make estimating C loss through soil CO2 efflux difficult. To determine the drivers of soil CO2 efflux, we selected six hardwood floodplain forests at the lower middle Elbe River, which were distributed among different relief positions (low‐lying or high‐elevated) in the active and former flooding zone. We measured soil CO2 effluxes over a full year using the closed‐chamber method. Based on the response of soil CO2 efflux to soil moisture and temperature, annual efflux rates were determined, which were then related to soil properties, such as pH, texture, soil organic carbon (SOC) and nitrogen (N) content. Soil CO2 efflux ranged between 1006 (±99) and 2214 (±118) gC m−2 year−1. Maximum efflux occurred in a former floodplain forest that was disconnected from Elbe River water table fluctuations. SOC‐specific soil CO2 efflux (gC gSOC−1 year−1) was smallest in low‐lying forests of the active flooding zone and reflected by the appearance of redoximorphic mottling close to the soil surface. Fine texture (<6.3 μm), SOC and N were related positively and electric conductivity, C/N and pH negatively to total soil CO2 efflux. Soil pH and fine texture were the strongest univariate predictors for total soil CO2 efflux (both R2 = 0.59). Fine texture, pH and C/N ratio explained 66% of the variance in total soil CO2 efflux according to multiple linear regression. We conclude that, in hardwood floodplain forests, soil CO2 efflux is mainly controlled by fine texture and soil pH. Fine texture can be related to soil moisture and nutrient availability and may have a positive effect on the activity of microorganisms.

• Maximum soil CO 2 efflux was measured in a disconnected floodplain forest • SOC-specific soil CO 2 efflux was smallest in forests where redoximorphic mottling occurred close to the surface • Fine texture (<6.3 μm) and pH were the strongest predictors for annual soil carbon cycle, climate change, global warming, heterotrophic and autotrophic respiration, hydromorphic features, riparian forest ecosystems, soil organic carbon, wetlands 1 | INTRODUCTION Floodplain ecosystems play a significant role in the global carbon (C) cycle (Battin et al., 2009;Cole et al., 2007). In these ecosystems, soil organic carbon (SOC) has been acknowledged to be the largest organic C pool (Sutfin et al., 2016). Natural hardwood floodplain forests have particularly strong SOC sequestration potential due to autochthonous and allochthonous inputs from vegetation and flood-induced sedimentation (Dybala et al., 2019;Lininger et al., 2018). C inputs are usually in an approximate balance with losses due to erosion and soil CO 2 effluxthrough autotrophic root respiration and heterotrophic soil organic matter decomposition in the litter layer and soil (Melack & Engle, 2009). However, today, this balance is often disrupted because forests are cleared, managed or situated behind a dike, that is, in the former flooding zone (Brunotte et al., 2009). Many hardwood floodplain forests have thereby lost their natural ecosystem functions, such as C retention, biodiversity and flood risk regulation (Hornung et al., 2019). Additionally, heterogenic relief and periodic flooding affect metabolic and biogeochemical processes dynamically (Wilson et al., 2011). Therefore, spatiotemporal dynamics of soil CO 2 effluxes can vary substantially in floodplain ecosystems.
The most important drivers of temporal changes in soil CO 2 efflux are soil temperature and moisture (TM) (Davidson et al., 1998). Soil CO 2 efflux from soil and litter is usually positively related to soil temperature and mainly caused by heterotrophic respiration, where its magnitude depends on characteristics of the microbial community composition, and availability of SOC and nutrients (Fang & Moncrieff, 2001). Seasonal changes in temperature affect autotrophic respiration due to C allocation patterns of trees, where C is directed downwards during the late phases of summer and directed upwards in shoot extension phases (Fan & Han, 2018). Soil CO 2 efflux is limited at high and low soil moisture (Tang & Baldocchi, 2005). Low soil moisture restricts the metabolic activity of roots and microbes due to a lack of soil water and dissolved organic matter. High soil moisture inhibits oxygen accumulation, caused by the water-filled pore space, which is needed for aerobic soil organic matter decomposition (Luo & Zhou, 2006). Due to inundation in floodplains, abrupt changes in soil moisture can cause shifts in microbial community structure and thereby affect soil respiration and enzymatic degradation rates (Wilson et al., 2011). Additionally, natural seasonal changes in soil moisture and temperature can have significant effects on soil CO 2 efflux (Janssens & Pilegaard, 2003).
In floodplains, microbial communities are adapted to relief-affected local moisture regimes, and the related variabilities of oxygen availability, soil pH, SOC and nutrient pool, and soil texture (Chen et al., 2014;Yin et al., 2019). Fine soil particles can physically protect SOC from decomposition by stabilisation within aggregates and organomineral associations (Deiss et al., 2017). These particles can also retain water and nutrients better than sandy soils, which are important for SOC decomposition (Hamarashid et al., 2010). Also, root respiration can be smaller in clayrich compared to sandy soils because low bulk densities lead to a smaller fine root density and autotrophic respiration decreases with root diameter (Tang et al., 2020). In floodplains, a micro-relief-driven heterogeneous soil texture distribution is very common (Drouin et al., 2011), which could have strong effects on the variability of soil CO 2 efflux. Sand-dominated soils are usually found on highly elevated embankments and loamy soils in low-lying relief positions (Schwartz et al., 2003). Due to these reliefaffected soil properties, soil CO 2 effluxes comparable to deserts (160 gC m À2 year À1 ) up to rates comparable to tropical forest (>1205 gC m À2 year À1 ) can occur in a single floodplain ecosystem (Doering et al., 2011).
In hardwood floodplain forests, metabolic and biogeochemical processes that contribute to soil CO 2 efflux are mainly controlled by dynamic changes in temperature, as well as relief, which affects soil moisture changes and sedimentation processes (Acosta et al., 2017;Chen et al., 2014;Cierjacks et al., 2010). These processes can alter the availability of SOC and nutrients, as well as pH, hydromorphic features (e.g., redoximorphic mottling) and soil texture (Heger et al., 2021). To understand the effect of these soil properties on soil CO 2 efflux in a central European floodplain ecosystem, we measured soil CO 2 efflux and its driving factors during 1 year in six hardwood floodplain forests along the lower middle Elbe River, differing in relief position (high and low) and flooding zone (active and former). We aim to (1) quantify the total annual soil CO 2 efflux, (2) characterise the effects of seasonal changes in soil TM, (3) and determine the effects of soil properties, such as pH, N and SOC content and texture, on soil CO 2 efflux.

| Site description and categorisation
The study area is located at the lower middle Elbe River in northern Germany ( Figure 1). We selected six hardwood floodplain forests dominated by Quercus robur (Table 1). The forests were between 18 and 186 years old (Shupe et al., 2021). Each study site within the forests covered an area of 625 m 2 . The study sites were selected avoiding border effects (e.g., anthropogenic disturbance, drainage) and had homogeneous relief, vegetation cover, and soil characteristics, respectively. The study sites were categorised by relief position into low-lying (low) and high-elevated (high) classes of the active and the former flooding zone of the Elbe River. The former flooding zone is situated behind a dike and active flooding cannot occur (Koenzen et al., 2021). The former low I (seepage) forest is hydrologically connected to the Elbe River through seepage water inflow, whereas the former low II (disconnected) forest is located adjacent to tributaries of the Elbe River (Sume-Konauer Graben and Sumter See), and is unaffected by Elbe River water table fluctuations.

| Assessment of soil properties
At each study site, three soil profiles-evenly distributed over the site-were characterised and sampled between September 2018 and May 2019. The reference soil groups were characterised after IUSS Working Group WRB (2015) (Table S1). Soil properties, such as SOC, nitrogen (N) content, their ratio (C/N ratio), pH (in 0.01 M CaCl 2 ), and texture contents (clay, silt, sand, and intermediate fractions), were determined per each soil horizon according to the methods described in Heger et al. (2021). Bulk means of the texture fractions, SOC, N, C/N ratio, and pH, and the sum of SOC stocks (t ha À1 ) were evaluated until 1 m below soil surface ( Table 2). The former low forest II is distinct from the other sites in texture class, with a fine texture content (sum of clay and fine silt, <6.3 μm, in %) of >50%, whereas active high forest I, is a rather sandy site with fine texture contents of <5%. All other sites have intermediate fine texture contents and are loamy soils. Electric conductivity (EC) was measured in a solid-to-solution ratio of 1:2.5. To study the effects of hydromorphic features, we recorded the depth up to 2 m below the soil surface of each soil profile with the first appearance of hydromorphic mottling, that is, where redoximorphic features covered an area of >5% of the soil horizon according to Ad-hoc-AG Boden (2005), referred to as "hydromorphy" (m). In soil profiles where we did not find hydromorphic features until 2 m, we set the depth to 2 m, assuming that processes that contribute to hydromorphic mottling start close to this depth.

| Site instrumentation
All soil profiles were equipped with volumetric water content and temperature sensors (CS650, Campbell Scientific, United Kingdom) at 10 cm depth. Hourly measurements were logged onto a data logger (CR300, Campbell Scientific, United Kingdom). At the active low II, active high I, active high II, and former low I forests, climate stations-consisting of air temperature and humidity sensors (CS215, Campbell Scientific) and a rain gauge (Kalyx-RG, Campbell Scientific)-were installed directly outside of the forest at each study site. At the former low I forest, the climate station was placed inside a clearing. Climate data from the active high II forest was regarded as representative of active low I and from the active high I for former low II due to their close proximity ( Figure 1). A data logger (CR1000, Campbell Scientific) recorded the data hourly. At the active high I forest, a barometer (Baro-Diver DI500, van Essen Instruments, The Netherlands) measured air pressure hourly. Groundwater fluctuations were measured hourly at every study site using a submersible transducer (TD-Diver, DI801, van Essen Instruments).

| Soil CO 2 measurements and efflux calculation
To estimate the CO 2 efflux from the soil surface, we used the closed-chamber method, where the rate of CO 2 concentration increase inside the chamber is used for CO 2 efflux evaluation (Madsen et al., 2010). To conduct repeated chamber measurements at the same location, we inserted cylindrical collars (PE-100, ThyssenKrupp Plastics GmbH, Germany)-with 15 cm height and 50 cm diameter-3 cm into the ground during January 2020. Three collars were inserted in a radius of 1-3 m from each of the three instrumented profiles at random locations so that each study site comprised nine chamber measurement locations in total. During measurement, a rubber ring was placed between the chamber and collar as a gas-tight seal.
The cylindrical chamber is made up of transparent acrylic glass (XT, 5 mm, Röhm GmbH, Evonik Industries, Germany) and has a dome-shaped top cover (acrylic glass 500 mm, HB Präsentationssysteme oHG, Germany). It has a diameter of 0.5 m and a volume of 0.12 m 3 . To conduct dark measurements, the chamber was wrapped in aluminium foil, which was covered with white duct tape to minimise radiative heating. A fan (120 mm, be quiet!-Pure Wings 2, Listan GmbH, Germany) mixed the air inside the chamber. A non-dispersive infrared (NDIR) CO 2 sensor module K30 (Senseair, Sweden) measured the CO 2 concentration inside the chamber. Air temperature and relative humidity were measured by a temperature and humidity sensor (Adafruit Sensiron SHT31-D, Adafruit Industries, United States) and air pressure by a barometric pressure and altitude sensor (Adafruit BMP280, Adafruit Industries) inside the chamber. All data was logged using a microcontroller (MEGA 2560 Rev3, Arduino.cc, Italy) equipped with a data logging shield (Adalogger FeatherWing, Adafruit). This logging device together with a 12 Volt battery was placed on the outside of the chamber. During measurements, the data were additionally displayed on a 4.5-inch display module (Joy-IT SBC-LCD20Â4, SIMAC Electronics GmbH, Germany), which was connected to and placed on top of the logging device.
The measuring period encompassed one full year of 366 days, from 27 February 2020 6:00 to 27 February 2021 5:00 (UTC). Soil chamber measurements were F I G U R E 1 Location of hardwood floodplain forest study sites. Low and high are height indications of the study sites. The forests are between 18 and 186 years old. The forests are either located in the active or former flooding zone of the Elbe River. The former low forest I is located behind a dike and affected by Elbe River water table fluctuations through seepage water inflow, from which the former low II (disconnected) is unaffected T A B L E 1 General study site characteristics: The relief positions of the study sites are height-indicated as low-lying (low) or highelevated (high) conducted at least once per month, with varying intervals per site. Field trips were partially restricted due to the CoVid-19-pandemic. Consequently, measurements during March and April 2020 as well as in December and January 2021 were not possible at some sites. Before the measuring campaign started, we clipped herbaceous vegetation and tree seedlings inside the collars to measure exclusively CO 2 effluxes from bare soil and litter. Any vegetation that grew inside the collars afterwards, was carefully clipped back prior to measurement. Chamberplacement-duration was 4 min per measurement. According to the ideal gas law, temperature and pressure affect the CO 2 readings of NDIR gas analysers in a stable volume. Furthermore, water vapour dilution affects the measurements; therefore, a correction was included in the calculations. The CO 2 sensor module K30 performs temperature compensation using a built-in temperature sensor. Soil air pressure correction and corrections due to the water vapour dilution effect on the measured CO 2 concentrations were applied according to the methods described in Heger et al. (2020). Soil CO 2 efflux (μmol m À2 s À1 ) was evaluated using the ideal gas law (Pérez-Priego et al., 2015): where (p À e a ) represents the partial pressure of dry air, e a is the partial pressure of water vapour, V C is the chamber volume plus the respective air volume for each collar above the soil surface, R is the ideal gas constant (8.314 J mol À1 K À1 ), T a is the air temperature, and A C is the soil area inside the collar. Δχc Δt is the CO 2 concentration gradient from 30 to 210 s after the chamber was placed onto the collar. We used linear regression (mean R 2 was 0.96 ± 0.06; median R 2 was 0.98) to evaluate the concentration gradient.

| Analyses of soil temperature and moisture response of CO 2 effluxes
To evaluate the soil temperature (T, in C) response of soil CO 2 effluxes, we used an exponential temperature model, with which we were also able to calculate Q 10 . Q 10 is a measure for the temperature sensitivity of soil CO 2 efflux per 10 C temperature increase (Lloyd & Taylor, 1994): Q 10 was calculated according to: where β 1 was derived from Equation (2). Equations with exponential expressions were log-transformed to estimate the coefficients by linear regression (cf. Tang and Baldocchi (2005)). To eliminate the effect of temperature on soil moisture (M, volumetric water content, in m 3 m À3 ), soil CO 2 efflux was normalised (Hirano et al., 2003).
where T b is the base temperature, which was set to 10 C. We then fitted a quadratic soil moisture function (Mielnick & Dugas, 2000): To model soil CO 2 efflux for up to 1 year, a combined TM model, which has been proven to perform best among forest ecosystems (Tang & Baldocchi, 2005;Webster et al., 2009;Yoon et al., 2014), was fitted: The linear regression R 2 , adjusted R 2 , akaike information criterion (AIC), and the mean square error (MSE) of a 10-fold cross-validation (Kohavi, 1995) were evaluated for Equation (6). Data gaps in hourly measured T or M (comprised in total <1% and occurred due to low battery capacity during winter) were filled using linear interpolation.

| Statistical analyses
We used Python v3.7.10 (Van Rossum & Drake, 2009) for one-way ANOVA (with post hoc tests), univariate linear regression analysis and (Pearson) correlation analyses. For ANOVA, Tukey HSD post hoc comparison was used to identify significant differences between floodplain forest categories. Variance homogeneity was tested using Levene's test (p < 0.05), and residuals were checked for normality using the Shapiro-Wilk test (p < 0.05). In case of variance inhomogeneity, a Welch-ANOVA was performed instead, and category comparisons were conducted using pairwise t tests with Bonferroni-Holm p-adjustment. Differences were considered significant for p < 0.05.
We used R v4.1.2 (R Core Team, 2021) for multiple linear regression analysis (MLR) and the Monte Carlo (MC) approach (using the MonteCarlo package). MLR was performed with a stepwise selection of predictors, based on the AIC. The model with the smallest AIC was selected. Residuals were checked for normality using the Shapiro-Wilk test (p < 0.05). Spearman correlation was used to identify correlations between the variables. Variance of inflation factor (VIF, with factors <2.3) was used to estimate collinearity between the predicting variables. To evaluate the uncertainties of the model used for annual soil CO 2 efflux evaluation (Equation 6), we applied the MC approach (Huang et al., 2020;McMurray et al., 2017). Ten thousand simulations were performed per each soil profile using the SEs of the model coefficients that we estimated by fitting Equation (6) to the measured data. Out of these annual soil CO 2 efflux simulations, a mean, a 95% CI, and SD was evaluated per soil profile.

| Seasonal course of soil CO 2 effluxes
Soil CO 2 effluxes were greatest during the summer months (June, July, and August) and smallest in winter (December, January, and February). During winter, soil CO 2 effluxes ranged between 0.18 μmol m À2 s À1 (in active high I) and 4.45 μmol m À2 s À1 (in former low II). Soil CO 2 effluxes during summer ranged between 2.05 (in active high I) and 15.88 μmol m À2 s À1 . Maximum soil CO 2 effluxes always occurred in former low II whereby maximum soil CO 2 effluxes of the other forests did not exceed 10.73 μmol m À2 s À1 . In former low II, the smallest measured soil CO 2 efflux during summer was 6.16 μmol m À2 s À1 . Soil CO 2 efflux among all study sites was larger during autumn (0.61-9.70 μmol m À2 s À1 ) compared to late spring (2.80-5.61 μmol m À2 s À1 ).
Mean air temperature ranged between 9.6 ± 7.8 C and 10.6 ± 8 C inside the forests. The mean annual precipitation ranged between 443.8 and 552 mm and was below the annual mean from 1991 to 2020 of Lüchow (Figure 1) of 563 mm (Deutscher Wetterdienst, 2022). There was no flooding during the study period and the closest measured groundwater table distance to the surface among the study sites was 82 cm (Figure 2).

| Effects of soil temperature and moisture on soil CO 2 efflux
The exponential T model (Equation 2) fitted the soil CO 2 efflux at every study site significantly (Figure 3). F I G U R E 2 Seasonal course of measured soil CO 2 effluxes during the study period in the six hardwood floodplain forests. Nine measurements per box. Due to the CoVid-19-measures, some sites during March and April 2020 as well as in December and January 2021 could not be monitored F I G U R E 4 Temperature-normalised soil CO 2 effluxes versus soil moisture at 10 cm for all study sites and soil profiles (P1, P2 and P3). *Significant fit of Equation (5) between x and y at p < 0.05 F I G U R E 3 Soil CO 2 effluxes versus soil temperature at 10 cm for all study sites and soil profiles (P1, P2 and P3). Error bars are SDs of the soil profiles (three spatial divided measurements per profile P1, P2 and P3). The R 2 and Q 10 were evaluated from the exponential fit of Equation (2). *Significant fit of Equation (2)   SDs of the measurements increased with soil temperature. The Q 10 ranged between 1.81 and 2.44 and was smallest in the former low I and greatest in the former low II forest. The Q 10 was not different among the study sites according to ANOVA.
F I G U R E 5 Annual soil CO 2 efflux for all six study sites. Low and high are height indications of the study sites. The forests are between 18 and 186 years old. The forests are either located in the active or former flooding zone of the Elbe River. The former low forest I is located behind a dike and affected by Elbe River water table fluctuations through seepage water inflow, from which the former low II (disconnected) is unaffected. Error bars indicate the SE of the mean (n = 3); significant differences according to ANOVA with Tukey post hoc comparison are annotated with letters F I G U R E 6 Annual total soil CO 2 efflux versus fine texture (<6.3 μm) (a), soil organic carbon (SOC) stocks (b), SOC content (c), N content (d), C/N ratio (e), soil pH (0.01 M CaCl 2 ) (f) and electric conductivity (EC) (g). The black line indicates linear regression between variables in x and y. *Significant linear regression between x and y at p < 0.05 F I G U R E 7 Annual soil organic carbon (SOC)-specific soil CO 2 efflux for all six study sites. Low and high are height indications of the study sites. The forests are between 18 and 186 years old. The forests are either located in the active or former flooding zone of the Elbe River. The former low forest I is located behind a dike and affected by Elbe River water table fluctuations through seepage water inflow, from which the former low II (disconnected) is unaffected. Error bars indicate the SE of the mean (n = 3); significant differences according to ANOVA with Tukey post hoc comparison are annotated with letters The quadratic M model-which describes an optimum at intermediate soil moisture-fitted at all measured locations in the active low I and active high II forest (Figure 4a,e). Optimum soil moisture for soil CO 2 efflux ranged between 0.17 and 0.26 m 3 m À3 in active low I, 0.22 and 0.28 m 3 m À3 in active low II, 0.09 and 0.11 m 3 m À3 in active high I, 0.10 and 0.16 m 3 m À3 in active high II, and between 0.05 and 0.11 m 3 m À3 in former low I. In the former low II forest, optimum moisture was 0.35 m 3 m À3 at P3, but no optimum moisture could be estimated for the other two soil profiles (Figure 3f). The measured soil moisture range was smallest at active high I and greatest in former low II with 0.01-0.17 m 3 m À3 and 0.22-0.48 m 3 m À3 , respectively. Soil moisture ranged between 0.06 and 0.41 m 3 m À3 in active low I, between 0.10 and 0.40 m 3 m À3 in active low II, between 0.04 and 0.29 m 3 m À3 in active high II, and between 0.02 and 0.31 m 3 m À3 in former low I.

| Differences between the forests
The TM model performances and the annual soil CO 2 efflux per study site are summarised in Table 3. The total soil CO 2 efflux differed between the low-lying forests of the active and the former flooding zone ( Figure 5). Annual soil CO 2 efflux in active low I was 38% smaller than in active low II and effluxes in former low I were 24% smaller than in former low I. The maximum total soil CO 2 efflux occurred in the former low II (disconnected) forest (Table 3) and was more than 36% larger than its counterparts in the active flooding zone.

| Soil properties controlling total soil CO 2 efflux
The strongest univariate predictors for annual total soil CO 2 effluxes were fine texture (positive relation; Figure 6a), and pH (negative relation; Figure 6f) (both R 2 = 0.59). Further positive relations of total soil CO 2 efflux were found to SOC, SOC stock and N (Figure 6b-d). A partial correlation analysis showed that the effects of SOC stocks were insignificant (p = 0.52) after removing the effects of fine texture, whereas the effects of fine texture stayed significant (p = 0.002) after removing the effects of SOC stocks on soil CO 2 efflux. EC and the C/N ratio had negative effects on total soil CO 2 efflux (Figure 6e,g).
In a stepwise MLR analysis, we used the same variables to explain variance in annual soil CO 2 effluxes, which have been used in these forests to predict SOC stocks (Heger et al., 2021): fine texture, soil pH, and C/N ratio. After the stepwise exclusion of confounding variables, the best-performing model included fine texture, pH and C/N ratio as predictors (Equation 7). They explained 66% of the variance in annual soil CO 2 effluxes (R 2 = 0.72, Adj. R 2 = 0.66, p < 0.001, AIC = 203.05). This shows that hardwood floodplain forests' soil CO 2 effluxes strongly relate positively to soil fine texture and negatively to pH and C/N ratio. Total soil CO 2 efflux ¼ 9:95 Â fine texture À 68:21 Â C=N À 298:70 Â pH 3.5 | Hydrologic control on SOC-specific soil CO 2 efflux To evaluate the SOC-specific soil CO 2 efflux we divided the total efflux by the SOC stocks ( Table 2). The SOC-specific soil CO 2 efflux of the active high I forest was more than twice as large as in its lowlying counterparts (Figure 7). The SOC-specific soil CO 2 efflux in active high I was 50% higher but not significantly different from its categorical partner forest (active high II) and from both forests in the former flooding zone. The SOC-specific soil CO 2 efflux seemed to be positively affected by hydromorphy ( Figure 8). Thus, a smaller proportion of C emitted in low forests where redoximorphic properties occur closer to the surface.

| Annual carbon budget of hardwood floodplain forests of the lower middle Elbe River
Total soil CO 2 efflux ranged between 1006 (±99) and 1625 (±135) gC m À2 year À1 in active floodplain forests and between 1688 (±87) gC m À2 year À1 and 2214 (±118) gC m À2 year À1 in former floodplain forests. Compared to similar ecosystems, total soil CO 2 effluxes of the current study were similar to an oak forest in northern Italy and to floodplain sites at the Danube River, where effluxes of up to 1800 (±621) gC m À2 year À1 have been measured (Ferréa et al., 2012;Schindlbacher et al., 2022). Mean soil CO 2 effluxes in the former low II forest ranging from 6.16 to 15.88 μmol m À2 s À1 during summer exceeded soil CO 2 effluxes in temperate hardwood floodplain forests in Czech Republic of 1.59-8.54 μmol m À2 s À1 (Acosta et al., 2017). The annual soil CO 2 efflux of 2214 (±118) gC m À2 year À1 in the former low II forests was also more than twice as high as in temperate biomes of 745 ± 421 gC m À2 year À1 (Bond-Lamberty & Thomson, 2010). One reason for these large effluxes could be the availability of a rich energy source for microbial decomposers at this site contained the largest litter layer C stock (Table 1). Also, no flooding conditions and a dry summer (mean annual precipitation was below the annual mean from 1991 to 2020 of Lüchow [ Figure 1] of 563 mm) (Deutscher Wetterdienst, 2022) could have contributed to large soil CO 2 effluxes among the studied forests.

| Seasonal effects of soil temperature and moisture on soil CO 2 efflux
The most important drivers for soil CO 2 efflux are soil TM. They control microbial and plant physiologic metabolism and diffusive flux in the air-filled pore space (Fan & Han, 2018). In our study, soil TM both had significant effects on soil CO 2 efflux (Figures 3 and 4).
The temporal course of soil CO 2 effluxes reached maximum efflux during summer and minimum during winter (Figure 2). The T model significantly explained the relationship between soil CO 2 efflux and temperature at all but one soil profile (Figure 3). It appears that at the sites with high soil CO 2 efflux (as in former low II), the data was less prone to scatter than at sites where efflux was less than <5 μmol m À2 s À1 , as in active low I P3. In general, however, the relationship between soil CO 2 efflux and temperature was exponential at all sites. This can be explained by the temperature sensitivity of root respiration, C oxidation by rhizosphere microbiota, as well as microbial soil organic matter decomposition . Greater effluxes in autumn compared to late spring (Figure 2) can also be affected by autotrophic respiration, which follows C allocation patterns of trees, where C is directed downwards during the late phases of summer and directed upwards in shoot extension phases (Fan & Han, 2018). Spatial heterogeneity from root and rhizosphere respiration and fine root production could also be responsible for the high SD with increasing soil temperature (Stoyan et al., 2000), because their sensitivity is larger during warmer periods . Thus, soil CO 2 efflux was positively related to seasonal changes in soil temperature among all studied forests.
Soil moisture affects soil CO 2 efflux directly by its effects on metabolic activity of roots and microorganisms and indirectly by providing available dissolved organic matter and oxygen for heterotrophic respiration. Usually small or large soil moisture levels limit soil CO 2 efflux (Tang & Baldocchi, 2005). The quadratic function (Equation 5) only described the data from two out of six forests of the current study (Figure 4) sufficiently, which could be due to small variability in soil moisture conditions during the measurement period. A small soil moisture range (Figure 4b,c,f) might have contributed to low model performance (except for active low II [ Figure 4d]). This effect could also have contributed to low TM model performances in active high I (R 2 = 0.29-0.62) and former low I (seepage) (R 2 = 0.38-0.50) ( Table 3). In floodplains, soil moisture is directly affected by periodic flooding (Yoon et al., 2014). However, flooding did not occur during the measurement period and flooding frequency is lower in highly elevated floodplains and in floodplains situated behind a dike compared to low-lying floodplains of the active flooding zone. The strong TM model performance of the second study site of the former flooding zone (former low II [disconnected]) could be related to the strong effect of temperature on soil CO 2 efflux (R 2 ≥ 0.75, Table 3). Thus, dry soil conditions and a lack of flooding could be responsible for the little effect strength of soil moisture on CO 2 effluxes in our forests.

| Effects of soil properties on soil CO 2 efflux
Spatial heterogeneity in soil CO 2 effluxes is strongly affected by the response of roots and microorganisms to soil texture, nutrient availability and pH (Yin et al., 2019). Annual soil CO 2 efflux was positively related to N, SOC stocks and fine texture. Soil CO 2 efflux was negatively related to pH, C/N ratio and EC. An MLR yielded a model that describes annual soil CO 2 efflux using the variables fine texture, C/N ratio and pH. We interpret these variables in a process-based way and argue that they reflect the positive effect of fine texture on water supply and available SOC as well as a microbial community, which is adapted to rather acidic conditions. Soil microbial communities are often adapted to specific pH and EC levels (Chen et al., 2014;Sardinha et al., 2003). The negative relation of soil CO 2 efflux to EC (Figure 6g) could reflect a stress effect on plants and microorganisms, which usually occurs in coastal wetlands (Song et al., 2021). The negative effect of pH on microbial activity tends to occur in alkaline soils (Yang et al., 2019). In low-saline floodplain soils with pH below 7, the contribution of fungi to the microbial community can be large (Sardinha et al., 2003). The soil pH of our study sites ranged between 4.3 and 6.3 (Figure 6f). Acidophilic bacterial and fungi species, which prefer this pH range, could have contributed to the negative relation of soil pH to CO 2 efflux (Luo & Zhou, 2006). A negative relation of soil CO 2 efflux to pH has also been found in Californian pine plantations (Xu & Qi, 2001), and in global annual soil CO 2 efflux models (Chen et al., 2014). In floodplain soils, a negative relation of pH with mineralization rate has been found when soil moisture was low but at an optimum level for microbial decomposition ($30% water holding capacity) (Yin et al., 2019). Considering that the experimental period of the current study comprised a year with MAP below the annual average (Deutscher Wetterdienst, 2022)-which was already followed by summer droughts in 2018 and 2019 (Boeing et al., 2021)-could indicate that the microbial community in our floodplain forests is adapted to these low pH and soil moisture conditions. Additionally, large EC can stress the microbial communities and plants and thus decline soil CO 2 efflux.
Soil microorganisms use organic matter-with its main components, SOC and N-as energy source for mineralization (Basile-Doelsch et al., 2020). This process could explain the positive relation of N and SOC stocks to soil CO 2 effluxes ( Figure 6). This relation could also explain the difference in annual soil CO 2 effluxes between forests of the same hydrologic situation (Figures 5 and 7). The C/N ratio indicates if the SOC pool is composed of plant detritus or more processed organic matter with a higher contribution of microbial residues (Schrumpf et al., 2013). In our study, C/N was negatively related to soil CO 2 efflux (Figure 6e), which could be related to a high contribution of microbial biomass at narrow C/N (Wagai et al., 2009).
Periodic flooding partly controls soil CO 2 efflux in wetland forests (Yoon et al., 2014), and oxygen scarcity in low-lying wetlands can reduce aerobic SOC mineralization (Yin et al., 2019). At Elbe River floodplains, groundwater fluctuations are controlled by the river at both sites of the dike, that is, floodplains in the seepage water zone are hydrologically connected to the Elbe River through seepage water inflow (Schwartz et al., 2003). This could explain why CO 2 effluxes in the floodplain forests of the seepage zone were similar to its active counterpart ( Figure 5). The positive correlation of hydromorphy with SOC-specific soil CO 2 efflux (Figure 8) also indicates SOC protection from mineralization due to saturation conditions close to the soil surface (Hennings et al., 2021). This relation was also reflected in the category comparison of the SOC-specific soil CO 2 efflux (Figure 7), where the low-lying forests in the active flooding zone experienced the smallest SOCspecific soil CO 2 efflux. Groundwater fluctuations can also control dissolution and precipitation of Fe and Mn, which can be involved in metal-organic complexes and protect SOC even under aerobic conditions (Wang et al., 2017). Thus, our study showed that in low-lying hardwood floodplain forests of the active flooding zone, SOC mineralization might be reduced due to oxygen scarcity, which was also reflected by the appearance of hydromorphic features close to the surface (Figure 8).
In low-lying hardwood floodplain forests, also fine texture can protect SOC from microbial degradation by adsorption to clay minerals or incorporation within clay and silt aggregates (Deiss et al., 2017). Soils with small fine texture content are less effective at retaining water with easily available dissolved organic matter and nutrients, which are important for mineralization (Hamarashid et al., 2010). This could explain the large SOC-specific effluxes in the sand-dominated active high I forest (Table 1 and Figure 7). However, the greatest total soil CO 2 effluxes were measured in the forest with maximum fine texture content (Table 1 and Figure 5). This could be attributed to a large availability of energy-rich labile SOC and N for mineralization, which can be associated with fine soil texture (Luo & Zhou, 2006). We also noticed soil dry cracks at former low II during summer, which additionally could have promoted the transport of CO 2 from deeper soil layers (Schindlbacher et al., 2022). Soil texture can also affect the root system, since compared to loamy soils, in sandy soils, low fertility, unsaturated hydraulic conductivity and water holding capacity negatively affect root vitality and decomposition of root litter, and thus autotrophic and heterotrophic respiration in the rhizosphere (Luo & Zhou, 2006). This effect supports the finding that autotrophic and heterotrophic respiration benefits from indirect positive effects of fine texture on soil moisture and availability of nutrients and SOC.
In the current study, soil CO 2 efflux-autotrophic and heterotrophic-could be related to soil properties, such as fine texture, its association with optimum moisture conditions, SOC and N availability, as well as to soil pH, and EC. The positive relation to fine texture and the negative relation to pH resulted to be the most important drivers for soil CO 2 efflux. However, autotrophic respiration can also be positively related to soil sand content and pH. Large sand content is related to soils with higher bulk densities where a denser fine root system is favourable, while respiration decreases with root diameter (Tang et al., 2020). High soil pH could have a positive feedback on forest primary productivity due to higher nutrient availability as a consequence of high soil cation exchange capacity (Härdtle et al., 2004), which would contribute to enhanced autotrophic respiration. In deciduous temperate hardwood forests, root and rhizosphere respiration can contribute up to 90% to total soil CO 2 efflux . No information about the contribution of root and rhizosphere respiration is known for hardwood floodplain forests, further studies are needed. Our results suggest that fine texture provides positive feedback mechanisms on soil moisture, SOC and nutrient stock, which is important for autotrophic and heterotrophic respiration. Low soil pH and EC could also be related to a well-adapted microbial community.

| CONCLUSION
Soil CO 2 efflux in Elbe River hardwood floodplain forests was mainly related to fine texture and soil pH. Seasonal temperature changes were related to effluxes in all studied forests. Soil moisture, as limiting factor for soil CO 2 efflux, was significantly related only in two out of six forests, which could be due to a lack of flooding conditions in the studied period. Total soil CO 2 efflux ranged between 1006 (±99) and 2214 (±118) gC m À2 year À1 . Maximum efflux occurred in former floodplain forests, which were disconnected from Elbe River water table fluctuations. Relief or flooding zone did not indicate differences in total soil CO 2 efflux between the other forests. However, relief effects on SOC-specific soil CO 2 efflux (gC gSOC À1 year À1 ) indicated that SOC might be protected from mineralization in lowlying forests of the active flooding zone due to oxygen scarcity. This protective effect on SOC mineralization was also reflected by the appearance of hydromorphic features close to the soil surface. Our study highlights the importance of relief-affected soil properties, such as soil pH and texture, for soil CO 2 efflux estimates and hardwood floodplain forest management.