We collected 10 soil samples from depths of 0–5 or 0–10 cm in six forests and a pastureland located in Niigata (six soils from four forests), Ibaraki (two soils from a forest) , and Oita (two soils from a forest and a pastureland) prefectures in Japan. Figure 1 and Table 1 present the locations and site characteristics, i.e., elevation, mean annual temperature (MAT), mean annual precipitation (MAP), potential evapotranspiration (PET), and net primary production (NPP). Briefly, all of the investigated sites have a humid temperate climate with MAT of 9.1–10.8 °C and MAP of 1474–2930 mm. All of the forests are dominated by beech (Fagus crenata and Fagus japonica) and oak (Quercus serrata), except for Oita forest which is a deciduous/evergreen mixed forest dominated by Siebold's maple (Acer sieboldianum), Japanese snowbell (Styrax japonicus), and Japanese holly (Ilex crenata). The pastureland in Oita is dominated by Japanese lawn grass (Zoysia japonica), dwarf fountain grass (Pennisetum alopecuroides), cranesbill (Geranium thunbergii), white clover (Trifolium repens), and Indian strawberry (Potentilla indica).

Soil sampling was conducted in the snow-free season (April to October) of 2021. We also collected soil samples from layers below the target depth (i.e., 0–5 or 0–10 cm depth) down to 50 cm as the maximum depth to examine whether the soil could be classified as Andisol. According to the USDA (United States Department of Agriculture) Soil Taxonomy criteria , soils with 60 % or more of the thickness containing more than 20 mg of acid-oxalate-extractable Al (Alo) and 1/2 Fe (Feo) per 1 g soil within a depth of 0–60 cm are classified as Andisols. According to these criteria, 4 of the 10 soils were Andisols (one each from Niigata and Ibaraki and two from Oita), with non-allophanic properties determined by a high ratio of pyrophosphate-extractable Al (Alp) content to Alo content (Alp/Alo>0.5; ; Fig. S1 in the Supplement). Collected soil samples were transferred to the laboratory and stored at 4 °C before further analysis. Before analysis, soils were gently passed through a 4 mm sieve to remove gravel and plant tissue. Fine roots in the sieved samples were removed with tweezers. In our study, we considered that the soil water content at the soil sampling reflected the ability of soil to hold the water and thus the usual water contents in the field because the soil water content showed significantly positive correlation with water-holding capacity (WHC) (r=0.87, p<0.01). Therefore, CO2 release rate for constant moisture conditions in the present study should represent the release rate under the usual field moisture conditions of each soil.

The soil properties analyzed were pH(H2O); electrical conductivity; water content; water-holding capacity (WHC); total carbon (C) and nitrogen (N) contents; particle size distributions as relative compositions of clay, silt, and sand-sized particles; and selectively dissolved Al and Fe contents (Tables 2–4). The pH(H2O) was measured in soil and water mixtures consisting of 1 g of soil and 2.5 mL of water. For measurement of electrical conductivity, 5 mL of water was added to 1 g of soil. Water content was measured by determining the difference in soil weight before and after drying at 105 °C for 24 h. WHC was measured as the difference in soil weight before and after water saturation referring to the Hilgard method . Here, water contents when soil is completely saturated in the Hilgard method should equal zero pF value (0 kPa) as soil water potential. Soil total C and N contents were measured for air-dried and well-ground soil samples using an elemental analyzer (vario PYRO cube; Elementar, Manchester, UK). Particle size distributions were determined using the sedimentation method based on Stokes' law using soil mineral particles after removing organic matter by hydrogen peroxide solution digestion. Selectively dissolved metals such as reactive Al and Fe were measured according to the procedure described previously in . Briefly, the contents of Al and Fe extractable with 2.0 M acid ammonium oxalate (i.e., Alo and Feo, respectively) were measured as contents of organo-metal complexes and short-range-order minerals, while Al and Fe extractable with 0.1 M sodium pyrophosphate solution (i.e., Alp and Fep, respectively) were measured as contents of organo-metal complexes . The difference between acid-oxalate- and pyrophosphate-extractable metals (i.e., Alo-p and Feo-p) represented the contents of short-range-order minerals . Contents of Al and Fe in the solution were measured with an inductively coupled plasma–optical emission spectrometer (ICP-OES) (5110; Agilent Technologies, Santa Clara, CA, USA). Carbon concentrations in pyrophosphate-extracted solution (Cp) were also measured with a total organic carbon (TOC) analyzer (TOC-L; Shimadzu, Kyoto, Japan). For soils from Kuju forest and grassland and two Ogawa forests, we also measured C concentration in free light density fraction (fLF) obtained by the density fractionation method using sodium polytungstate solution at 2.0 gcm-3 , in order to evaluate possible effects of labile C abundance in fLF on CO2 release increase by DWCs.

For the soils after incubation (see Sect. 2.3), soil microbial biomass C and N were measured by the chloroform fumigation extraction method . Organic C and total N concentrations in 0.5 M potassium sulfate solution used for extraction of fumigated, and nonfumigated soils were measured using another TOC analyzer (TOC-L; Shimadzu) equipped with a total nitrogen (TN) unit (TNM-L, Shimadzu).

Soils were incubated aerobically at 20 °C for 84 d including three DWCs (i.e., 28 d per cycle). Simultaneously, soils were incubated in the same manner but without DWCs, during which water content of the soils was maintained at a constant level equivalent to the mean water content for the DWC treatment (Fig. 2). A pre-incubation was conducted at the constant water content for 7 d prior to the 84 d incubation. A post-incubation was also conducted at the constant water content for 28 d after the 84 d incubation to evaluate the remaining effect of DWCs on soil CO2 release. We consider that the incubated soils have been aerobic even after the rewetting to increase the water content by twice the WHC because the CO2 concentrations in our experiment never surpassed 1 %; thus the oxygen concentrations in the incubation jar have likely never decreased below 19 % or lower. Also, a sufficiently large volume of our incubation jar (1.0 L) compared to contained soil amounts (i.e., 5.31–10.63 g) and added water contents in the rewetting (i.e., ca. 6 to 7 mL) support the state of aerobic condition during the incubation.

Mason jars (1.0 L volume; Ball, Buffalo, NY, USA) with lids equipped with tube fitting systems for gas sample collection were used as incubation jars. Small vials (300 mL, SM sample glass vial; Sansyo, Tokyo, Japan) containing 5.31–10.63 g of soil sample depending on water content were placed in mason jars. Under the constant water treatment, we surrounded the small vial with 20 mL of water within the incubation jar to prevent the soil from drying. For each DWC, day 1 to day 7 and day 18 to day 24 were drying stages (Fig. 2), during which the soils were incubated with silica gel (20 gjar-1), which lowered the water content to <5 % WHC by day 7. Day 8 to day 12, day 13 to day 17, and day 25 to day 28 were the driest, wettest, and moderately wet stages, respectively, and CO2 release rates in these three stages were measured using a gas chromatograph equipped with a thermal conductivity detector (GC-14B; Shimadzu). The CO2 release rates were also measured in the pre- and post-incubation periods. At the beginning of the wettest stage, soils were rapidly rewetted with distilled water to double the soil water content from the initial status of DWC incubation. For CO2 release rate measurements, at the start of each stage, the headspace of the incubation jar was flushed with CO2-free air for 15 min at a rate of 0.5 Lmin-1, and the jar was closed. At the end of each stage, 15 mL of gas sample was collected from the jar using a 20 mL plastic syringe (Terumo, Kyoto, Japan) and stored in a pre-evacuated 5 mL glass vial (SVG-5; Nichiden Rika, Osaka, Japan). Then, the CO2 release rate was determined from the increase in CO2 concentration during this period. After gas sampling, the jars were flushed with CO2-free air and closed for the next incubation stage. For 84 d in the incubations with the constant moisture condition as controls, the CO2 release rates were measured for day 1 to day 12, day 13 to day 28, day 29 to day 40, day 41 to day 56, day 57 to day 68, and day 69 to day 84, in addition to the pre- and post-incubation periods. Soil water contents during the incubation were measured periodically and maintained by adding water to ensure the same mean water content between the two treatments. Even in the drying stage under the DWC treatment for day 1 to day 7 and day 18 to day 24, we conducted measurements of soil water content once or twice. The measurements were performed by weighing those soils. Based on these data, we confirmed that the mean soil water content during DWC incubation was equal to that during constant moisture incubation. All incubations were conducted with three replicates for each treatment and soil.

The CO2 release rates were compared between the DWC and constant water content treatments. For the DWC treatment, the CO2 release rates in the drying stages (i.e., day 1 to day 7 and day 18 to day 24) could not be measured and therefore had to be estimated to evaluate the mean CO2 release rates for the individual cycle and total of three cycles. The rates in the drying stages were estimated as the mean values of the CO2 release rates measured before and after the period of interest, in the same manner as described previously . Then, the effect size of the DWCs on CO2 release (defined here as the increase factor, IF_CO2) was quantified as the ratio of CO2 release rate under the DWC condition to that under the constant water content condition . Here, doubling of CO2 release by DWC resulted in IF_CO2 of 2, while halving resulted in IF_CO2 of 0.5.

The pairwise t test was applied to examine the statistical significance of differences in CO2 release rates between the DWC and constant water content treatments with adjustment of site-by-site variations in the metrics. Differences in soil microbial biomass C and N between the two treatments were also evaluated with the pairwise t test. To explore predictors explaining the variation in IF_CO2 among soils, relations between IF_CO2 and environmental and soil properties were visualized on scatterplots and evaluated by linear correlation analysis. All statistical analyses were conducted with R 4.1.1 (R Core Team, Vienna, Austria), and p<0.05 was taken to indicate statistical significance.

The CO2 release rates under DWC conditions showed large fluctuations for all soils, along with fluctuations in soil water content (Fig. 3). The CO2 release rates in the driest stages (3.5 %–18.2 % of WHC depending on the soil) were 3.0–41.5 µgC per gram dry soil per day. In contrast, CO2 release rates in the wettest stages (90.8 %–201.1 % of WHC) reached 18.8–194.1 µgC per gram dry soil per day. In the moderately wet stages, CO2 release rates were 8.4–63.9 µgC per gram dry soil per day. For the constant water content treatment, CO2 release rates observed for each soil showed little fluctuation during incubation although the rates varied from 7.3 to 69.9 µgC per gram dry soil per day depending on the soil.

For all soils, the observed large fluctuations in CO2 release under DWC treatment resulted in a greater CO2 release rate than under constant water content treatment, although the IF_CO2 values varied among the soils (Fig. 4). At the wettest stage in the first DWC, the CO2 release rates increased by 47.2–127.7 µgC per gram dry soil per day compared with those under the constant water content conditions, resulting in IF_CO2 values of 2.9–12.2. For the whole of the first cycle (i.e., the first 28 d), the CO2 release rates increased by 15.4–43.4 µgC per gram dry soil per day under the DWCs compared with the constant water content conditions, resulting in IF_CO2 values of 1.6–5.2. For the whole incubation period (84 d) including three DWCs, IF_CO2 values were 1.3–3.7, with an increase in CO2 release rate by 7.4–23.8 µgC per gram dry soil per day by DWCs. These increases in CO2 release by DWCs were observed in all cycles during the 84 d incubation period, whereas no increase was observed in the 28 d post-incubation period after the three DWCs (Fig. S2 in the Supplement). Considering these results, we focused on the IF_CO2 variations obtained for the whole incubation period, including three DWCs.

Among the environmental and soil physiochemical properties, reactive mineral and metal contents (especially, Alo + 0.5Feo, Alo, Feo, Alp + 0.5Fep, and Alp) in soils showed significant positive correlations with IF_CO2 (p<0.05; Table 5). In particular, Alp content appeared to be a key predictor of the variation in IF_CO2 among the soils, given that Alp accounted for most (73 %–99 %) of Alo (Table 4) and showed a higher correlation coefficient than others (Table 5). Scatterplots for Alp content and IF_CO2 values are presented in Fig. 5. In addition, the molar ratios of soil total C and Cp contents to Alp contents showed significant negative correlations to IF_CO2 (p<0.05; Table 5, Fig. 5). Here, it should be noted that IF_CO2 had no significant relationship with soil water content at soil sampling and WHC, suggesting the variation in IF_CO2 among soils resulting from other than such hydrogenic properties of soils. The SOM quality, such as C/N ratios of total bulk soils and K2SO4-extractable fractions, also had no significant relationship with IF_CO2. The insignificant relations of SOM quality to IF_CO2 were also supported by the fact that Kuju forest and grassland soils having more than doubled differences in fLF abundance showed only 12 % differences in IF_CO2, and two Ogawa forest soils having almost identical fLF abundances showed more than doubled differences in IF_CO2. The amounts of clay and sand-sized particles showed significant correlations with IF_CO2 after rewetting in the first cycle (p<0.05, r=-0.66 for clay and 0.71 for sand particles). However, those correlations between the particle contents and IF_CO2 were insignificant for IF_CO2 for a total of three cycles. Thus, in the present study, reactive mineral and metal content, especially Alp content in soils, was the significant predictor for IF_CO2 variation among soils, rather than soil water environments, SOM quality, and particle size distributions.

Among the additional soil properties related to soil microbial activity and abundance, the soil-C-content-specific CO2 release rate under constant moisture conditions (defined as the CO2 release rate per unit of C in soil, qCO2_soc, in µgCO2-C per milligram of soil carbon per day) showed significant negative correlations with IF_CO2 values in all incubation stages (p<0.01; Table 6, Fig. 6). Here qCO2_soc should be an index for microbial availability of carbon substrate normalized by total C contents in soils. Thus, using qCO2_soc, we can consider whether the microbially available carbon substrate in interested soil should be much more than in other soils. Microbial biomass N also showed significantly negative correlations with IF_CO2 values after rewetting in the first cycle and in the whole of the first cycle (p<0.05) but not in the whole incubation period including three cycles (p=0.12).

There were also considerable relations among Alp contents, total C/Alp molar ratio, and qCO2_soc (Fig. 7). Soil Alp content and total C/Alp molar ratio showed a negative correlation at p<0.05. The qCO2_soc showed a positive correlation with total C/Alp molar ratio and negative correlation with Alp content at p<0.01. The Cp/Alp molar ratio also showed similar covariations with these variables, showing a strongly positive correlation with total C/Alp molar ratio (r=0.90, p<0.01; Fig. S3 in the Supplement).

It should be noted that both microbial biomass C and N were significantly lower in soils incubated under DWC conditions than constant water content conditions (p<0.05; Fig. 8). Microbial biomass C under DWC conditions was lower by ≥20.2 % than that under constant water content conditions. Microbial biomass N under DWC conditions was ≥12.6 % lower than under constant water content conditions. Nevertheless, there were no significant correlations between the differences in microbial biomass and the value of IF_CO2 (p≥0.15; Fig. 9).