4.1 Effect of rainfall on soil moisture dynamics
In the Chinese Loess Plateau (CLP), precipitation is the main source of soil moisture (SM) (Jia et al., 2017; Su and Shangguan, 2019). Although the average annual precipitation decreased over nearly half a century on the CLP (Fu et al., 2017), the season-rainstorm occurrence frequency did not significantly change with vegetation construction (Tang et al., 2018). However, under rainstorms, the runoff (Lü et al., 2012) and sediment (Wang et al., 2016) concentrations decreased significantly after afforestation, such as the “Grain For Green Project”, radically altering the rainfall-SM response relationship, which intercepts rainfall, delays surface runoff, and increases rainwater infiltration (Kijowska-Strugala et al., 2018). Nevertheless, the characteristics of the soil water cycle process have changed, and the extent and process of this change are still unclear.
Our results concluded that the rainfall-SM response exhibited a positive but not synchronous correlation. The fluctuation of SM occurred after precipitation, especially in larger amounts of rainfall. The shallow soil layers were more susceptible to rainfall than deep layers (Fig. 2). This phenomenon was more obvious under different rainfall patterns with different rainfall amounts, durations, and intensities in the CLP (Hou et al., 2018; Jin et al., 2018; Tang et al., 2019). For example, heavy rains (Group Ⅰ) with a larger amount and intensity had the deepest SM response depth, at least 70 cm depth (Fig. 3), which was similar to Wang et al. (2013). Tang et al. (2019) also illustrated that a larger rainfall amount could promote rainwater percolation into deeper soil. However, high intensity with a continuous input of rainfall might surpass the max-rate of soil permeability and limit the soil water infiltration in the shallow layer (Yan et al., 2021), resulting in the shortest lasting time of the 10-cm depth and the slowest permeating velocity of the 100-cm profile (Tables 3 and 4). Compared with heavy rains, intermediate rains (Group Ⅱ) showed the shallowest response depth (Fig. 3) but the shortest response time (RT) of the entire profile and the fastest wetting front velocity (WFV) of the 10-cm SM (Tables 3 and 4). It was indicated that small rains with a high rainfall intensity could penetrate the dense canopy cover and litter layer to trigger a surface SM response in the most effective way of all the rainfall patterns, which was similar to the results of Liu et al. (2020). Continuous rains (Group Ⅳ) were the most lagging and slowest pattern to cause an SM feedback. For instance, continuous rains not only lagged in RT of the entire depth but also showed the smallest WFV in the 1-m profile across all rainfall patterns. This result was mainly due to a smaller rainfall amount with a longer duration stretching the rainfall time and decreasing the average rainfall intensity, which made it difficult to through the canopy and litter layer, and to store rainwater in the surface soil layer. There has not been enough water to trigger the SM response and infiltrate into the deep layer at a larger speed of the soil wetting front, due to the lack of guidance from the gravitational water potential (Mao et al., 2018). Therefore, a high average rainfall intensity with a smaller rainfall amount promotes a quick surface SM response, but a larger rainfall amount facilitates rainwater percolation into deeper soil with a larger WFV.
In addition, regardless of the vegetation pattern, the surface SM (10-cm depth) responded only when the minimum accumulated rainfall amount (ARA) surpassed 5 mm. This conclusion is slightly different from Jin et al. (2018), who demonstrated that a 9 mm ARA was necessary to trigger surface SM variation. Perhaps the difference in canopy coverage and litter depth due to vegetation age, which revegetation after 60 years in Jin et al. (2018). After 20 years of revegetation in this region, a 5 mm ARA was the threshold needed to replenish soil water, which was consumed by plant-inducing dried soil layers (Jia et al., 2017; Wang et al., 2011). It is suggested that the constraint of a minimum ARA exists in terms of vegetation restoration in semiarid CLP.