Multiple abiotic and biotic drivers of soil water storage capacity in temperate forests recovering from disturbances

Soil water storage capacity acts as a vital forest function to intercept rainfall and retain water for plant growth processes. However, whether or how plant functional trait diversity and composition regulate soil water storage capacity remains poorly understood. Structural equation modeling (SEM) was used to detect the direct and indirect effects of multiple biotic (i.e., functional trait composition and functional diversity) and abiotic (topography and soil organic carbon) factors on soil water storage capacity, i.e., in terms of soil capillary water storage content (CW), soil non-capillary water storage content (NCW), and soil saturated water storage content (TSW), in temperate forests recovering from different logging disturbance intensity levels.


Introduction
Soil water storage capacity is one of the major soil functions which could potentially affect biogeochemical cycle, and hence, mitigate the in uences of global climate change (Lü et al. 2015;Reich et al. 2018; Zhang et al. 2016). For example, soil water storage capacity regulates the water ow to alleviate the negative effect of hydrological variations, which in turn could regulate climate change events (Andreassian 2004; Zhang et al. 2011). In addition, the plant diversity effect usually presented a signi cant relation with soil water-related functions, indicating that plant diversity is one of the main biotic factors underlying soil water storage capacity (Kammer et al. 2013;Wen et al. 2019). Although several studies have reported the differences of soil water storage capacity among diverse vegetation types (Fang et al. 2011;Hao and Wang 1998;Jiao et al. 2011;Wang et al. 2013), none of the studies has reported on the complex interactions of biotic (functional trait diversity and community-weighted mean trait composition; CWM) and abiotic (i.e., topography, climate, and edaphic) factors for determining soil water storage capacity in forests recovering from logging disturbances (see Fig. S1 for a conceptual model).
It is generally well explored that the niche complementarity (the diversity effect) and mass ratio (the CWM effect) mechanisms could explain several ecosystems functions in natural assembled communities (van der Plas 2019). For example, in the semi-arid grassland ecosystem of China, plant diversity is positively correlated with soil water content with the increasing community coverage (Wu et al. 2014). In addition, higher species diversity could improve soil water in ltration capacity by changing soil properties as the long-term feed-back between soil and high above-and below-ground biomass ). Indeed, the ecological mechanism behind this positive effect might be associated with the niche complementarity hypothesis i.e., species having different niches can use the available resources or facilitate their coexistence, then could improve ecosystem functions (Lange et al. 2019;Tilman et al. 1997). Besides the diversity effect, the trait of most dominant species (i.e. measured in CWM) could also drive ecosystem functions under the assumption of the mass ratio hypothesis (Grime 1998). For instance, the CWM of speci c leaf area (CWM SLA ) could explain more variance of water retention capacity than other predictors in semi-arid forest ecosystems (Teixeira et al. 2020). Moreover, tree species functional group, litterfall quality per se, and its decomposing rate rather than the diverse or the amount of litterfall input better explained the variance of soil functions (Dawud et al. 2017). Taken together, these prominent ecological theories could provide the quantitative relationships between diverse plant species, composition and soil water storage capacity, and hence may improve our understanding regarding underlying mechanisms when simultaneously testing their effects on ecosystem functions ( Many advanced ecological studies have suggested that abiotic factors (e.g., topography) can potentially affect plant diversity, composition and ecosystem functions (Jucker et al. 2018). For example, changes in soil physicochemical properties could be highly attributed to topographical factors (Chapin et al. 2011;Zhang et al. 2019). Fine soil particles, as well as soil organic carbon, tended to accumulate in lower elevational gradients because of the soil erosion processes when rainfall events appeared (Chapin et al. 2011;Zhang et al. 2019). Accordingly, soil water retention ability could be enhanced by ne soil texture, whereas soil water in ltration function could be improved by increased soil organic carbon contents (Chen et al. 2019;Fischer et al. 2015;Zhang et al. 2019). On the other hand, elevation as an important topographical factor can regulate soil water storage capacity indirectly through in uencing climatic factors which in turn could impact vegetation structure and composition (Hao et al. 2003). For example, along with the decline of elevational gradients, plant composition presents to be a quick return on investment species, and this kind of litterfall could be bene cial for the accumulation of soil organic carbon which in turn could in uence the soil structure and water storage capacity (Cotrufo et al. 2013;Fischer et al. 2019).
Under global change circumstances, forest ecosystems are impacted by a diversity of human activities, i.e., logging disturbance which is a common disturbance for timber use in forests (Edwards et al. 2014), resulting in changing vegetation structure and functions. The disturbance activities in forests could lead to alteration of forest types, associated above-and below-ground functions (Millar and Stephenson 2015;Yuan et al. 2018) as well as impact biogeochemical processes (de Avila et al. 2018). As such, the time lag effect of soil water storage functions may appear because of the different recovering rate of ecosystem functions and the process of plant-soil feedback (Thom and Seidl 2016;Trumbore et al. 2015). For instance, a decade-long Jena Experiment proved that plant species diversity affecting soil water contents from the shading effect in the early years was shifted to the soil properties in later years (Fischer et al. 2019). Although the effects of different types of disturbance (e.g., logging, re, cutting, clearing) on different forest functions have been widely reported (Edwards et al. 2014), there is no actual study dealing with soil water storage capacity in the forests recovering from logging disturbances. This knowledge is crucial for understanding forests resilience and recovering from disturbances as increasingly considered vital in the face of climate change and human interventions ).
This study aims to explore how multiple biotic (i.e., functional trait composition and functional diversity) and abiotic (topography and soil organic carbon) factors drive soil water storage capacity, i.e., in terms of capillary water storage content (CW), soil non-capillary water storage content (NCW), and soil saturated water storage content (TSW), in temperate forests recovering from different logging disturbance intensity levels ( Fig. 1). To do so, we used data from eleven permanent forest sites (in total 260 subplots) on Changbai Mountain in the northeast of China where forests underwent through three disturbance intensity levels. Using the conceptual model ( Fig. 1), we answered the following main research questions. 1) Which ecological mechanism -the mass ratio or the niche complementarity -explain variation in soil water storage capacity? 2) How do disturbance and topographic factors affect soil water storage capacity directly, and indirectly via functional trait diversity, CWM of trait, and soil organic carbon. 3) How does soil organic carbon mediate the effects of disturbance, elevation, functional trait diversity, and CWM of trait on soil water storage capacity? We hypothesize that multiple abiotic and biotic factors jointly regulate soil water storage capacity in temperate forests recovering from logging disturbances.

Study area, sites and forests
Our study area is located in the northeast of China, covering the Liaoning and Jilin Provinces (Fig. 1), which is classi ed as a temperate continental climate. The mean annual air temperate is 2.8°C, and the coldest and warmest monthly mean are − 13.7°C in January and 19.7°C in July (Hao et al. 2007). The mean annual participation is 700 mm which mostly occurs during the growing seasons (from June to September). Broad-leaved Korean pine (P. koraiensis) mixed forest, well-known for their high species diversity, productivity and complex stand structure, was the dominant vegetation type in the study region.
The predominant soil type of the studied area is Eutric cambisols according to the FAO soil classi cation system (Hao et al. 2007). Some areas had been subjected to variable intensities of human disturbances, but there were no severe human disturbances since 1998 when the forest protection policy was strictly implemented (Dai et al. 2004). Thus, forests recovering from disturbances include stands with different successional stages in the study area (Chen et al. 2014).
To quantify the effects of abiotic factors, disturbance intensity, plant trait diversity and composition on soil water storage capacity, we established 11 permanent forest sites during 2012 ~ 2013 (Table 1), and then re-investigated those forest plots at an interval of ve years using the standard eld approaches (Yuan et al. 2018). Within each site, all trees with a diameter at breast height (DBH) > 1 cm were marked, measured, recognized to species level, and positioned in contiguous 20×20 subplots (Hao et al. 2007). In this way, our data covered 22,766 stems in total belonging to 81 species, 46 genera, and 26 families across 260 subplots (Yuan et al. 2018). In addition, the topography of each subplot was measured by assessing the elevation of four corners of each subplot using Electronic Distance Measuring Device (Hao et al. 2007), and then the mean elevation, convexity and slope were calculated for each subplot (Harms et al. 2001). Elevation of the studied plots ranged from 640.4 to 1023.1 m.  Collectively, plots were primarily classi ed into three disturbance intensity levels according to the partial harvesting (e.g., thinning, selective harvesting): relatively low (< 10%), medium (10-20%), and high (20 ~ 30%) disturbance. Plots with medium and high disturbance levels were primarily located around the residential area, whereas plots with low disturbance level were located in the main region of the Changbai Mountain Nature Reserve (Fig. 1), which was established in 1960 and is part of the World Biosphere Reserve Network under the Man and the Biosphere Project in 1980 (Shao et al. 1994).

Quanti cation of plant functional trait diversity and composition indices
For quantifying the multiple facets of plant diversity, six functional traits were measured which were closely associated with forest growth, recruitment and death (Yuan et

Quanti cation of soil water storage capacity and soil organic carbon
In 2018, we randomly selected three soil sampling sites in the midpoints between the central point and four corners in each 20 m×20 m subplot. In each sampling site, two soil corers using stainless cylinders of 100 cm 3 in volume were selected for the bulk soil density and capillary water storage contents measurement after removing large debris. The corers containing large roots were abandoned for the precise analyses of data. Subsequently, ve soil cores (3.8 cm in diameter, 10 cm deep) at each sampling point were collected, pooled, and transferred to the laboratory with plastic zipper bags for further chemical analyses. Each soil sample was further divided into two parts, i.e., one for soil organic carbon analysis using the dichromate oxidation method (Lu 1999), and another one for soil moisture measurement after 12 hr dried at 105°C.
Soil water storage capacity was measured through soil porosity, and this could be divided into capillary porosity, non-capillary porosity and total porosity corresponding to soil capillary storage (CW), soil non- Total soil porosity (TP) was measured based on the measured soil bulk density while assuming that soil particle density is 2.65g cm -3 , W C was additional water weight after placing the stainless cylinders soil core in a tray with a 5-mm level of water until lter paper at the top of each core became moist (Liu et al. 2009), whereas non-capillary was the difference between TP and CP (Eq. (3)) (Wu et al. 2016).

Statistical analyses
We assessed the effects of disturbance on soil water storage capacity (i.e. TSW, NCW and CW) using a two-way ANOVA with Tukey's test as a post hoc analysis to assess the signi cant differences among disturbance levels. As such, we also tested the differences for associated variables, which may in uence soil water storage capacity (i.e. above-ground and below-ground variables), among disturbance intensity levels.
Before testing the conceptual model in Fig. S1, we assessed the spatial autocorrelation in the response variables (TSW, NCW and CW) among subplots using the generalized least-square models (GLS) with and without spherical autocorrelation. Our GLS analysis indicated that the models without spherical autocorrelation structures usually showed the lower Akaike Information Criterion (AIC) values compared to spherical autocorrelation models (Table S2, S3 and S4), suggesting that there was no strong spatial autocorrelation among subplots.
Based on Pearson's correlation (Table S5), the best combination of variables (Tables S6 and S7), and the conceptual model (Fig. S1), we identi ed elevation, CWM of traits, functional trait diversity, and soil organic carbon as the best factors in uencing soil water storage capacity (i.e. CW, NCW, and TSW). To test the conceptual model in Fig. S1, we used structural equation modeling (SEM) because it allows us to test the multiple theories, direct and indirect effects. To critically evaluate the best-tted SEM (Hoyle 2012), we used the highest Bentler's Comparative Fit Index (CFI > 0.90), the root mean square error of approximation (RMSEA ≤ 0.05), chi-square (X 2 ) test (P-value > 0.05), standardized root mean square residual (SRMR ≤ 0.05) and lowest AIC value. To compare the mass ratio hypothesis and niche complementarity effects on soil water storage capacity, we selected CWM SLA and FEve as the representative of plant functional trait composition and diversity, respectively, in SEMs. The direct, indirect, and total effects of predictors on response variables were calculated. The relative importance of each predictor was calculated as the percent of the given predictor standardized coe cient to the sum of standardized coe cients of all predictors (Yuan et al. 2019).
To complement the results from SEMs, we also conducted partial regressions and simple bivariate regressions (Grace et al. 2016). The SEM analysis was conducted using the "lavaan" package (Rosseel 2012). All analyses were conducted in R. 4.0.2 (Team RDC, 2019). All predictors in our research were standardized to a mean of 0 and standard deviation of 1, and the response factors (i.e. CW, NCW and TSW) were natural-log transformed before further analyses. Summary of variables used in the analysis is provided in Table S1. 3. Results

ANOVA: differences for soil water storage capacity and associated variables among disturbance intensity levels
The ANOVA results showed that CW and TSW signi cantly increased with increasing disturbance intensity, but no signi cant differences were found for NCW (Fig. 2). The average values for CW, NCW, and TSW of the top 20 cm soil layer were respectively 1118.2 t hm − 2 , 468.2 t hm − 2 , and 1604.7 t hm − 2 while showing general variations across eleven forest plots and disturbance intensity levels (Fig. S2). Meanwhile, above-ground biomass and soil bulk density decreased whereas plant species diversity indices and soil organic carbon increased with increasing disturbance intensity levels (Fig. S3). Furthermore, disturbance also severely in uenced most of the community trait composition; for example, CWM LA , CWM SLA , CWM LNC , and CWM LPC increased whereas CWM MH and CWM LCC decreased with increasing disturbance intensity levels (Fig. S3).

SEMs: what determines soil water storage capacity
The tested best-tted SEMs showed that FEev did not signi cantly in uence soil water storage capacity (i.e. CW, NCW, and TSW) whereas CWM SLA in uenced them either directly and/ or indirectly via soil organic carbon ( Fig. 3; Tables S8, S9 and S10). More speci cally, CWM SLA had signi cantly increased CW but decreased NCW directly. However, CWM SLA improved NCW and TSW indirectly via soil organic carbon ( Fig. 3; Tables S8 and S9). Although disturbance did not in uence soil water storage capacity directly, it had indirect variable (positive or negative) effects on soil water storage capacity via promoting CWM SLA and soil organic carbon, respectively ( Fig. 3; Tables S8, S9 and S10). Likewise, elevation showed indirect effects on soil water storage capacity via decreasing CWM SLA and soil organic carbon, but it directly increased NCW and TSW, and decreased CW ( Fig. 3; Tables S8, S9 and S10). In SEMs, soil organic carbon demonstrated the signi cant importance of mediating the effects of elevation, disturbance and CWM SLA on soil water storage capacity, as it had crucial improvement in uence on NCW and TSW directly( Fig. 3; Tables S8, S9 and S10).
The relative contribution analysis (Fig. 4a) showed that elevation was the most important in uencing factor for CW but not for TSW, whereas elevation, soil organic carbon and disturbance contributed much to the explained variation in NCW. Disturbance and soil organic carbon seemed to be the most important in uencing factors for TSW. However, FEve and CWM SLA seemed to be of additional importance for explaining variation in CW, NCW and TSW, and were almost equally important across CW, NCW and TSW (Fig. 4a). In addition, the comparison of direct and indirect effects showed that the direct effects of disturbance, elevation, soil organic carbon, FEve and CWM SLA were much higher than indirect effects on CW (Fig. 4b). However, the indirect effects of disturbance mattered for explaining variation in NCW, whereas the indirect effects of disturbance, elevation, CWM SLA and FEve mattered much for explaining variation in TSW (Fig. 4b).

Supporting analyses: partial regressions and simple bivariate regressions
The SEMs results showed that soil organic carbon was the important variable which in uenced soil water storage capacity directly as well as mediated the effects of disturbance, elevation and CWM SLA (Fig. 3).
We, therefore, further conducted the simple regression and partial regression analyses between soil organic carbon and soil water storage capacity (Fig. S4) to clarify the importance of soil organic carbon to soil water storage capacity. The partial regression analysis showed that soil organic carbon was signi cantly positively related with NCW and TSW but not related with CW when other factors (disturbance, elevation, CWM SLA , and FE VE ) were kept constant (Fig. S4a). This was a contrast with the simple regression analysis (Fig. S4b), which showed that soil organic carbon was signi cantly related with CW but not related with NCW, whereas soil organic carbon showed a signi cant constant relationship with TSW in both simple and partial regression analyses (Fig. S4). The partial regression results were in line with SEMs results, indicated that soil organic carbon was controlled by elevation, disturbance and CWM SLA , which in turn in uence NCW and TSW but not CW (Fig. 3 and Fig. S4a).
All bivariate relationships for hypothesized paths in SEMs are provided in Fig S4, S5

Discussion
In this study, we tease apart the direct and indirect effects of abiotic (elevation and soil organic carbon) and biotic (functional trait diversity and composition) on soil water storage capacity in temperate forests recovering from logging disturbances. Our results can be discussed in the following points: 1) CWM of a trait (functional composition) was a much more important direct factor to soil water storage capacity than functional trait diversity as well as providing an important role for mediating the effects of elevation and disturbances on soil water storage capacity; 2) disturbance in uenced soil water storage capacity indirectly via soil organic carbon and CWM of trait but not directly; 3) soil organic carbon was also the key factor in uencing soil water storage capacity directly as well provided an important role for mediating the effects of elevation, disturbances and CWM of trait on soil water storage capacity; and 4) elevation regulated soil water storage capacity directly and indirectly via soil organic carbon and CWM of a trait.
First, we found that plant community trait composition rather than functional trait diversity was the key Second, we observed that logging disturbance signi cantly changed community trait composition (Fig.  S3), and then changed soil water storage capacity in studied forests (Fig. 3)

Conclusions
Logging disturbances and elevation modulate the effects of CWM of traits on soil water storage capacity, and the quick return on investments of CWM of speci c leaf area shows a positive effect on soil water storage capacity (CW and TSW). Hence, CWM of speci c leaf area is a key biotic predictor for explaining variation in soil water storage capacity compared to functional trait diversity, supporting the mass ratio mechanism in temperate forests recovering from disturbances. As such, soil organic carbon was of additional importance to soil water storage capacity by affecting it directly or providing mediator role to link the responses of soil water storage capacity to elevation, disturbances, and CWM of speci c leaf area. We argue that testing the effects of multiple abiotic and biotic factors on soil water storage capacity could advance our understandings into the in uential ecological mechanisms underlying soil water storage capacity in forests under the context of human in uences.

Figure 4
Relative contribution of each predictors (a), and its direct and indirect effects (b) on soil water storage capacity based on best-tted SEMs. The solid color lled bar plots (b) shows direct effect whereas the pattern color lled bar shows indirect effect.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Thesupportingtables.xlsx Thesupportingtext.docx