The study was conducted in the Mao’er Mountain (110°20°- 110°35°E, 25°48°-25°58°N) (Fig. 1) area in the upper reaches of the Lijiang River basin, southern China. The source of the Lijiang River is in the Mao’er Mountain area with a mean elevation of 2100 m. This area has a subtropical monsoon climate affected by winter monsoon or westerly circulation transporting continental air masses and warm moist air from the western Pacific Ocean. Mean annual temperature is 7 ℃, and the highest air temperature is 23 ℃ in June and the lowest air temperature is -19 ℃ in January. Mean annual precipitation is 2100 mm, with approximately 76% of annual precipitation falling from the months of March to August. The majority of rainfall events are short and low intensity. Forest coverage is 96.5% and the dominant forests are bamboo forest, evergreen broadleaf forest, evergreen coniferous and broadleaf mixed forest, evergreen deciduous and broadleaf mixed forest, evergreen coniferous and broadleaf mixed forest, alpine forest and alpine meadow from low elevation to high elevation.

2.2.1 Experiment design

We selected three typical forests with tree layers composed mainly of P. pubescens forest (PPF), S. superba forest (SSF) and C. lanceolate forest (CLF). We established two plots in each forest type. Each plot was a plastic board-sided box 20 m long, 10 m wide, and 1 m deep, with a laterally continuous reinforced plastic pipe on the side and bottom to flux soil water. In mountainous regions, topography is a dominant factor in the deep soil layer when lateral water flux shapes the spatial distribution of soil moisture (Brocca et al., 2009). All of experimental plots in this study were, therefore, located on slopes with this characteristic. Variations of forest vertical structure and soil properties of different plots seemed to play a major role in influencing the spatial pattern of soil moisture. Basic information of the three typical forests is shown in Table 1.

2.2.2 Soil sampling and analysis

Three soil sampling points were designated by drawing an ‘S’ shape within each plot. Undisturbed soil samples were collected from three depths at each point in each plot. A soil auger was used to sample soils from 0-20, 20-50 and 50-80 cm layers in August 2013. Soil characteristics of the samples were measured in a laboratory using routine methods. Soil bulk density was estimated using a standard method (ISSCAS, 1978; Blake and Hartge, 1986). Total porosity and capillary porosity were determined by a method of ISSCAS (1978) and Liu et al. (2009). Non-capillary porosity was obtained by calculating the difference between total porosity and capillary porosity. Soil infiltration capacity was investigated using the double ring method (Ma et al., 2006). Litterfall within one-meter square in the plots were collected into plastic bags, transported to the laboratory and dried naturally. After drying, maximum water holding capacity was determined by wetting the material to saturation and then comparing difference in weight before and after wetting (Wohlfahrt et al., 2006).

2.2.3 Soil moisture testing and data collection

An Environmental Integration System was installed at each observation site to monitor climatic conditions at 5-min intervals, including air temperature, relative humidity, and precipitation during the study period (He et al., 2012). The probes for a soil moisture observation system (SMR101A-5, MadgeTech, America) were buried in the soil at desired depths with a cable extending to the surface to record soil moisture, soil temperature, maximum and minimum temperature in soil layers at 0-20, 20-50 and 50-80 cm at 5-min sampling intervals from 11 June 2013 to 15 October 2014.

2.2.4 Statistical analysis

The variances in soil moisture and soil properties between different soil layers and forest types were compared by one-way ANOVA, LSD’s test, and t test. The correlation between soil moisture and soil characteristics was analyzed by Spearman correlation analysis. All statistical analyses were conducted using SPSS 16.0 software (SPSS Inc., Windows ver. 16.0).

Table 2 summarizes the key physical properties from litterfall layer to deep layer, including litterfall, maximum water holding capacity of litterfall, soil bulk density, cap-illary porosity, non-capillary porosity, total porosity, initial infiltration rate, and steady infiltration rate. Many of these properties differed significantly among the three forest types. Litterfall differed with the mean values of 11.21 t ha^-1,14.73 t ha^-1, and 13.49 t ha^-1 for PPF, SSF and CLF, respec-tively. Litterfall in SSF was significantly higher than that in PPF and CLF; therefore, maximum water holding capacity of litterfall in SSF (32.8 t ha^-1) was significantly greater than in PPF (13.8 t ha^-1) and CLF (25.7 t ha^-1) (Table 2).Bulk density and porosity are important parameters reflect-ing permeability of water and the extension of plant roots (Zheng et al. 1998). The soil bulk density varied significantly in both soil layers and forest types. Soil bulk density increased with increasing soil depth (Table 2). From top layer (0-20 cm) to deep layer (50-80 cm), the soil bulk density increased from 0.85 g cm^-3 to 1.09 g cm^-3, 0.74 g cm^-3 to 0.88 g cm^-3, and 0.79 g cm^-3 to 0.92 g cm^-3 in PPF, SSF and CLF, respectively. Soil total porosity is composed of soil capillary porosity and soil non-capillary porosity. Soil porosity reflects the permeability of air and water, as well as the limit of plant root growth (Liu and Huang 2005). Results of the present study (Table 2) showed significant differences in soil porosity between soil layers in three typical forests. With increasing depth in the soil, the highest values of soil total porosity were observed at 0-20 cm soil layer, 0-20 cm soil layer and 20-50 cm soil layer for PPF, SSF and CLF, respectively. Total porosity in 0-20 cm soil layer was significantly higher in SSF (62.4%) than it was in PPF (57.70%) and CLF (50.21%). Infiltration features reflect the ability of soil to transform surface runoff into interflow or groundwater runoff and obviously affect soil moisture and water conservation (Wu et al. 2004). Results of this study showed no significant difference in the saturated permeability of soil upper and lower strata between the three different forests (Table 2). Most of the initial infiltration rate decreased with increasing depth in the soil, except for PPF, which increased slightly in 50-80 cm soil layer (Table 2).

Soil moisture in 0-20 cm, 20-50 cm and 50-80 cm layerswas monitored and analyzed in three typical forests. Theresults showed that soil moisture properties varied with periods (Fig. 2). During the study period, soil moisture in PPF ranged from 38.01 to 41.19%, from 39.04 to 42.78% and from 38.26 to 50.07% in 0-20 cm soil layer, 20-50 cm soil layer and 50-80 cm soil layer, respectively, from 47.55 to 53.84%, from 41.08 to 49.37% and from 42.42 to 48.39% in SSF, and from 35.01 to 38.61%, from 30.61 to 33.97% and from 27.60 to 34.51% in CLF. As expected, the three typical forests had the same change trends in response to rainfall. Soil moisture variations responded well to rainfall and most of the high values for soil moisture emerged in March, April or May during the year of observations. Soil moisture peaked in rainy seasons and then decreased in dry seasons. Although the precipitation was lowest in June, soil moisture was not lowest because soils retained the water obtained during rainy seasons (Fig. 2). The highest values in PPF were 40.33%, 29.29% and 25.95% in 0-20 cm layer, 20-50 cm layer and 50-80 cm layer, respectively, 39.34%, 25.50% and 22.58% in SSF, respectively, and 39.34%, 25.50% and 22.58% in CLF, respectively.

Vertical variations of mean monthly soil moisture at 0-20 cm, 20-50 cm and 50-80 cm depths over one-year observation period are presented in Table 3. The results show that soil moisture varied with forest types and soil layers. Soil moisture was lowest (39.71%) at 0-20 cm, and increased to 41.06% at 20-50 cm and to 42.97% at 50-80 cm in PPF, suggesting that soil moisture was mainly sourced from the precipitation infiltration. However, similar patterns were not observed in the other two forest types, SSF and CLF. The high value of soil moisture in SSF was 50.53% at 0-20 cm and the lowest value (43.70%) was found at 20-50 cm, while in CLF the highest value was 36.75% at 0-20 cm and the lowest value was 31.47% at 50-80 cm. The difference of soil moistures at 0-20 cm, 20-50 cm and 50-80 cm depths among three different structures of forest were significant, not only at 0-20 cm (P < 0.01), but also at 20-50 cm (P < 0.01) and 50-80 cm (P < 0.01). In the 0-20 cm layer, the SSF had the highest soil moisture of 50.53%, followed by PPF (39.71%) and CLF (36.75 %) (Table 3).

An analysis was carried out to understand how soil moisture varies with rainfall and soil properties (Tables 4 and 5). The monthly variations of soil moisture were closely related to variations of precipitation at the sites. A high positive correlation was found between precipitation and soil moisture, which proves that rainfall events play a critical role in raising soil moisture in three typical forests. Table 5 shows correlations among soil moisture and soil physical properties, which indicate that soil moisture was greatly affected by capillary porosity (R=0.983, P<0.05), non-capillary porosity (R=0.976, P<0.05), and total porosity (R=0.998, P<0.05). However, correlation analysis showed there was a negative relationship between soil moisture and initial infiltration rate (R=-0.985, P<0.05). Further analysis clearly showed that the soil initial infiltration rate had a positive correlation with non-capillary porosity (R=-0.946, P<0.05), which is consistent with the findings reported by Sharda and Ojaswi (2006). In addition, soil bulk density was significantly correlated with the litterfall (R=-0.983, P<0.05), implying that litterfall may be an important variable in controlling soil bulk density. However, no significant correlations between porosity and litterfall were observed in this study (Table 5).

Soil moisture variation was found in different forest types and soil layers. It is well known that organic matter strongly affects the physical and chemical characteristics of topsoil (Zhang et al., 2010). Since litterfall in SSF was higher than it was in PPF and CLF, the 0-20 cm soil layer has higher organic matter content and soil moisture in SSF than in PPF and CLF (Table 2). Of all the forest stands, the lowest soil bulk density was observed in the SSF, followed by PPF and CLF, which demonstrates that broadleaf forests are more efficient than coniferous forests in improving soil physical properties (Ma et al., 2006). In general, organic matter content has significant positive correlation with total porosity, capillary porosity, and non-capillary porosity, and significant negative correlation with bulk density; these correlations were observed in our study (Table 5). Hu et al. (2003) reported that forest vegetation could improve forest soil porosity and water holding capacity, and enhance the moisture-retaining, infiltration and water conservation functions. Hudson (1994) also found that for each one percent increase in soil organic matter, the available water holding capacity in the soil increased by 3.7%. Similar results were found by Wang et al. (2008), who found that the topsoil layer under high vegetation cover, as with the highly organic fine-grained soils and litter layer, reduced thermal conductivity and increased water infiltration and water hold capacity. Thus, we can conclude that bulk density, capillary capacity and porosity are the key factors affecting the physical and chemical properties of soil. These factors have important functions improving the physical and chemical properties of soil as well as advancing the cycle of soil water in the forest ecosystem.

As mentioned above, changes in soil properties may significantly influence soil infiltration. Under similar climatic and topographic conditions, soil infiltration is mainly controlled by soil properties. Soil infiltration increases with increasing soil sand content, soil porosity and soil organic matter, and decreases with increasing soil bulk density, soil clay content and initial soil water content (Neris et al., 2012; Zhao et al., 2014). Furthermore, our results suggest that soil infiltration may be affected by root expansion. Plants modify soil structure mainly when their roots grow into dense soil layers, and the amount of extension, death and decomposition of plant roots can benefit soil melioration by increasing soil aggregate and improving the soil porosity condition. Plant roots play an important role in soil moisture redistribution (Neumann and Cardon, 2012). In this study, PPF is a shallow root species while SSF and CLF are deep-rooted species. The slightly higher moisture at 50-80 cm in three typical forests (Table 3) might be the result high porosity in SSF and CLF (Table 2), while it may be the result of high lateral flow within the soil layer in PPF. Unfortunately, because of the complexity of root systems in forest systems, the spatial heterogeneity of root biomass distribution was not determined in this study.

Gaining a better understanding of the interactions between vegetation, soil and water flux is essential to environmental management for enhancing water yield in water-limited environments. Soil water is one of the most important factors controlling hydrological processes (Castillo et al., 2003; Seeger et al., 2004). However, soil moisture is influenced by vegetation structure and soil properties. Vertical structural factors such as tree rooting density and tree canopy can influence the spatial and temporal variability of soil moisture (Staelens et al., 2006; Dalsgaard, 2007). Forest soils tend to be relatively porous and accumulate more organic matter with high infiltration rates because trees loosen the soil. As presented in Table 3, the observation sites in each typical forest exhibited large variations in soil moisture both in space and depth. Significant correlations between soil moisture, precipitation and soil physical properties were observed in this study (Table 4 and 5). Soil moisture changed with rainfall, suggesting that soil moisture was mainly affected by precipitation (Table 4). We found significant Perrson’s correlations (P<0.01) between soil litterfall and soil properties, which confirmed that soil litterfall strongly controls soil properties, especially in the top soil (Zhang et al., 2010). This is supported by the observations reported by Venkatesh et al. (2010). Comparison showed that the soil moisture at 0-20 cm depth was higher in SSF than in PPF and CLF. A possible reason for this difference lies in the fact that the SSF site had very high organic matter content that improved the soil moisture of the soil (Zhao et al., 2006). In the deep layer, there were significant relationships between soil moisture and porosity and infiltration rate in the three typical forests, which was mostly because of the distribution of roots in the soil layers. Much of the soil moisture heterogeneity was resulted from variations in preferential water flow related to root distribution. Moreover, the generation of macrospores and channels by root penetration through soil tends to form preferential flow paths, thus enhancing soil moisture in deep soil (Li et al., 2011; Benegas et al., 2014). In brief, soil moisture was affected by various factors, including topography, precipitation, soil properties and vegetation. Many studies have pointed out that it is difficult to identify the relative contributions of vegetation, soil properties and topography to variations of soil moisture because of strong mutual and multiple influences among these influencing factors (Penna et al., 2009; Wang et al., 2012). All of these factors influence the spatial and temporal variability of preferential water flows in the soil, and on the other hand, bring significant variability in soil moisture regimes (Zheng et al., 2008).