Sustainability of Abandoned Slopes in the Hill and Gully Loess Plateau Region Considering Deep Soil Water

Soil desiccation of the deep soil layer is considered one of the main limiting factors to achieving sustainable development of ecosystems in the hill and gully Loess Plateau region. In this study, slope croplands were selected as the control, and deep soil water was studied on abandoned slopes, including natural abandoned slopes, Robinia pseudoacacia plantations, and Caragana korshinskii plantations. Then, we explored deep soil water characteristics of different vegetation types and slope aspects and the variation tendencies of deep soil water at different recovery stages. The results showed that there were no significant differences in deep soil water content between sunny and shady slopes, and thus, slope aspect was not the key impact factor affecting deep soil water. Deep soil water content on R. pseudoacacia plantations and C. korshinskii plantations was lower than that on natural abandoned slopes; there were no significant differences in soil water content between the natural abandoned slopes and slope croplands. Soil desiccation did not exist on natural abandoned slopes; thus, natural vegetation restoration is an appropriate way to achieve a sustainable ecosystem with respect to deep soil water. In contrast, soil desiccation intensified until it was difficult for vegetation to obtain available water in the deep soil layer on the plantations; soil desiccation began to appear at the 11–20-year stage, and it became increasingly severe until the deep soil water was close to the wilting coefficient at the ≥30-year stage on R. pseudoacacia plantations. Deep soil water was rapidly consumed, and soil desiccation began to appear at the 1–10-year stage and then was close to the wilting coefficient in the later stages on C. korshinskii plantations. According to the results, the plantations needed to be managed in a timely manner to prevent or reduce soil desiccation.


Introduction
The ecological environment is fragile due to long-term cultivation and destruction of vegetation in the hill and gully Loess Plateau region [1]. Vegetation recovery of slope cropland is an essential way to improve the ecological environment, and the main restoration approaches are abandonment and forestation [2,3]. Relevant research demonstrates that natural vegetation is suitable to the ecological environment in the hill and gully Loess Plateau region [4]; for the plantations, as invader plants, Robinia pseudoacacia and Caragana korshinskii have advantages in terms of vegetation recovery because of their resistance to drought and infertility [5]. Thus, under the vegetation restoration project of China's government, a large area of slope croplands has been converted into grasslands, shrublands, or woodlands during the past few decades, and natural abandoned slope, R. pseudoacacia plantation, Zhangjiahe, Sanwanggou, Zhifanggou, and Xiannangou watersheds were selected from north to south in the county (Figure 1). These watersheds are located in a forest-steppe region. Natural forest is almost absent as a result of long-term human activities that destroyed the natural vegetation and of farms on slope lands [25]. However, the slope cropland was gradually abandoned, and vegetation began to be restored in the past decades, especially since 1999.
A large number of typical slopes were widely distributed in these watersheds and they were selected for the investigation and sampling. As shown in Table 1, natural abandoned slopes (87 slopes), R. pseudoacacia plantations (65 slopes), and C. korshinskii plantations (10 slopes) were selected as the study subjects, and slope croplands (6 slopes) were selected as controls. Natural abandoned slopes, R. pseudoacacia plantations, and slope croplands were divided into sunny natural abandoned slopes (45 slopes), shady natural abandoned slopes (42 slopes), a sunny R. pseudoacacia plantation (33 slopes), a shady R. pseudoacacia plantation (32 slopes), sunny slope croplands (3 slopes), and shady slope croplands (3 slopes). The C. korshinskii plantations were not divided into sunny and shady slopes because they were mainly distributed on top of a hill. According to the years of restoration, the slopes were divided into four groups: 1-10-year stage, 11-20-year stage, 21-30-year stage, and ≥30-year stage (Table 1). Zhangjiahe, Sanwanggou, Zhifanggou, and Xiannangou watersheds were selected from north to south in the county (Figure 1). These watersheds are located in a forest-steppe region. Natural forest is almost absent as a result of long-term human activities that destroyed the natural vegetation and of farms on slope lands [25]. However, the slope cropland was gradually abandoned, and vegetation began to be restored in the past decades, especially since 1999.
A large number of typical slopes were widely distributed in these watersheds and they were selected for the investigation and sampling. As shown in Table 1, natural abandoned slopes (87 slopes), R. pseudoacacia plantations (65 slopes), and C. korshinskii plantations (10 slopes) were selected as the study subjects, and slope croplands (6 slopes) were selected as controls. Natural abandoned slopes, R. pseudoacacia plantations, and slope croplands were divided into sunny natural abandoned slopes (45 slopes), shady natural abandoned slopes (42 slopes), a sunny R. pseudoacacia plantation (33 slopes), a shady R. pseudoacacia plantation (32 slopes), sunny slope croplands (3 slopes), and shady slope croplands (3 slopes). The C. korshinskii plantations were not divided into sunny and shady slopes because they were mainly distributed on top of a hill. According to the years of restoration, the slopes were divided into four groups: 1-10-year stage, 11-20-year stage, 21-30-year stage, and ≥30-year stage ( Table 1).  Figure 2). Deep soil water was not influenced by annual precipitation, generally [13]; however, 2013 was an extreme flow year (Figure 2), and rainfall infiltration depths could reach the 500 cm soil layer and 300 cm soil layer on natural abandoned slopes and tree plantations, respectively [26]. Therefore, we removed the data from 2013 for error reduction, and the data for soil water were used from the years 2003-2007, 2012, and 2015.
Two sampling points were selected on the upper and lower part of each slope, respectively, to measure soil water in July or August. The samples were collected using a soil auger (10 cm in diameter) to a depth of 500 cm and taken at 20 cm intervals. The samples were oven dried at 105 • C for 24 h in the laboratory, and then, the weights of the dried samples were measured. Then, the mass water content was obtained.  Figure 2). Deep soil water was not influenced by annual precipitation, generally [13]; however, 2013 was an extreme flow year (Figure 2), and rainfall infiltration depths could reach the 500 cm soil layer and 300 cm soil layer on natural abandoned slopes and tree plantations, respectively [26]. Therefore, we removed the data from 2013 for error reduction, and the data for soil water were used from the years 2003-2007, 2012, and 2015.
Two sampling points were selected on the upper and lower part of each slope, respectively, to measure soil water in July or August. The samples were collected using a soil auger (10 cm in diameter) to a depth of 500 cm and taken at 20 cm intervals. The samples were oven dried at 105 °C for 24 h in the laboratory, and then, the weights of the dried samples were measured. Then, the mass water content was obtained.

Statistical Analysis
Mean soil bulk density was 1.20 g cm −3 on the study slopes, and thus, volumetric water content was calculated through the mass water content and soil bulk density. The soil water content of a slope in the 200-500 cm soil layer was the average of the soil water contents taken at 20 cm intervals in the 200-500 cm soil layer. Then, the mean soil water content could be obtained from natural abandoned slopes, R. pseudoacacia plantations, C. korshinskii plantations, and so on.
According to the definition of dried soil layer, a dried soil layer was present when soil water content was lower than the stable field capacity [27,28]. Field capacities and wilting coefficients were determined through the observation of 27 slopes, including natural abandoned slopes, R. pseudoacacia plantations, C. korshinskii plantations, and slope croplands in the study area. The mean field capacity was 0.19 cm 3 cm −3 . The mean wilting coefficient was 0.06 cm 3 cm −3 . The stable field capacity was 0.11 cm 3 cm −3 , which was calculated based on the field capacity multiplied by 60% [29]. Available soil water storage was calculated as follows: where Aw, Sw, Wc, and Th are available soil water storage (mm), volumetric water content (cm 3 cm −3 ), wilting coefficient (cm 3 cm −3 ), and thickness of soil layer (mm).
Differences in soil water content among the slopes of the studied vegetation types and control slopes were examined with one-way ANOVA. Differences in soil water content among different recovery stages were also examined with one-way ANOVA. The values were transformed using log (x + 1) to meet the homogeneity of variance assumption.

Statistical Analysis
Mean soil bulk density was 1.20 g cm −3 on the study slopes, and thus, volumetric water content was calculated through the mass water content and soil bulk density. The soil water content of a slope in the 200-500 cm soil layer was the average of the soil water contents taken at 20 cm intervals in the 200-500 cm soil layer. Then, the mean soil water content could be obtained from natural abandoned slopes, R. pseudoacacia plantations, C. korshinskii plantations, and so on.
According to the definition of dried soil layer, a dried soil layer was present when soil water content was lower than the stable field capacity [27,28]. Field capacities and wilting coefficients were determined through the observation of 27 slopes, including natural abandoned slopes, R. pseudoacacia plantations, C. korshinskii plantations, and slope croplands in the study area. The mean field capacity was 0.19 cm 3 cm −3 . The mean wilting coefficient was 0.06 cm 3 cm −3 . The stable field capacity was 0.11 cm 3 cm −3 , which was calculated based on the field capacity multiplied by 60% [29]. Available soil water storage was calculated as follows: where Aw, Sw, Wc, and Th are available soil water storage (mm), volumetric water content (cm 3 cm −3 ), wilting coefficient (cm 3 cm −3 ), and thickness of soil layer (mm).
Differences in soil water content among the slopes of the studied vegetation types and control slopes were examined with one-way ANOVA. Differences in soil water content among different recovery stages were also examined with one-way ANOVA. The values were transformed using log (x + 1) to meet the homogeneity of variance assumption.

Deep Soil Water in Different Vegetation Types and Slope Aspects
Mean volumetric water content was 0.16, 0.15, 0.10, and 0.07 cm 3 cm −3 at a depth of 200-500 cm on slope croplands, natural abandoned slopes, R. pseudoacacia plantations, and C. korshinskii plantations, respectively, and accordingly available soil water storages were 286, 257, 112, and 34 mm, respectively. As shown in Figure 3A, the distribution of water content was 0.06-0.21 cm 3 cm −3 on abandoned slopes, and they were 0.09-0.21, 0.06-0.17, and 0.06-0.08 cm 3 cm −3 on natural abandoned slopes, R. pseudoacacia plantations, and C. korshinskii plantations, respectively. The water content of R. pseudoacacia plantations and C. korshinskii plantations were significantly lower than that of the others (p < 0.05). There were no significant differences in water content between slope croplands and natural abandoned slopes, and between R. pseudoacacia plantations and C. korshinskii plantations ( Figure 3A), and there were no significant differences in the water content between sunny and shady slopes on natural abandoned slopes and R. pseudoacacia plantations ( Figure 3B,C).

Deep Soil Water in Different Vegetation Types and Slope Aspects
Mean volumetric water content was 0.16, 0.15, 0.10, and 0.07 cm 3 cm −3 at a depth of 200-500 cm on slope croplands, natural abandoned slopes, R. pseudoacacia plantations, and C. korshinskii plantations, respectively, and accordingly available soil water storages were 286, 257, 112, and 34 mm, respectively. As shown in Figure 3A, the distribution of water content was 0.06-0.21 cm 3 cm −3 on abandoned slopes, and they were 0.09-0.21, 0.06-0.17, and 0.06-0.08 cm 3 cm −3 on natural abandoned slopes, R. pseudoacacia plantations, and C. korshinskii plantations, respectively. The water content of R. pseudoacacia plantations and C. korshinskii plantations were significantly lower than that of the others (p < 0.05). There were no significant differences in water content between slope croplands and natural abandoned slopes, and between R. pseudoacacia plantations and C. korshinskii plantations ( Figure 3A), and there were no significant differences in the water content between sunny and shady slopes on natural abandoned slopes and R. pseudoacacia plantations ( Figure 3B,C). , and on natural abandoned slopes (B) and R. pseudoacacia plantations (C) with slope aspects. Note: different letters denote significant differences at the 0.05 level among different sample types. SCL, NAS, RPP, CKP, SuA, ShA, SuR, and ShR indicate slope cropland, natural abandoned slope, R. pseudoacacia plantation, C. korshinskii plantation, sunny natural abandoned slope, shady natural abandoned slope, sunny R. pseudoacacia plantation, and shady R. pseudoacacia plantation, respectively. The median is indicated by the horizontal line in the boxes, and solid boxes include the first and third quartiles. Whiskers extend to farthest points that are not outliers.

Deep Soil Water at Different Stages on Natural Abandoned Slopes
As shown in Figure 4, water content did not significantly decrease in the 300-400 cm and 400-500 cm soil layers on sunny slopes, nor in the 200-300 cm, 300-400 cm, and 400-500 cm soil layers on shady slopes in comparison to that on slope croplands at the 1-10 years, 11-20 years, 21-30 years, and ≥30-year stages. Water content in the 200-300 cm soil layer during the ≥30-year stage was smaller than that at the 1-10-year stage on sunny slopes ( Figure 4A). There were no dried soil layers on the sunny and shady slopes ( Figure 5). In addition, the water content of the sunny slopes was significantly lower than that of shady slopes in the 340-500 cm soil layer in the ≥30-year stage (p < 0.05). , and on natural abandoned slopes (B) and R. pseudoacacia plantations (C) with slope aspects. Note: different letters denote significant differences at the 0.05 level among different sample types. SCL, NAS, RPP, CKP, SuA, ShA, SuR, and ShR indicate slope cropland, natural abandoned slope, R. pseudoacacia plantation, C. korshinskii plantation, sunny natural abandoned slope, shady natural abandoned slope, sunny R. pseudoacacia plantation, and shady R. pseudoacacia plantation, respectively. The median is indicated by the horizontal line in the boxes, and solid boxes include the first and third quartiles. Whiskers extend to farthest points that are not outliers.

Deep Soil Water at Different Stages on Natural Abandoned Slopes
As shown in Figure 4, water content did not significantly decrease in the 300-400 cm and 400-500 cm soil layers on sunny slopes, nor in the 200-300 cm, 300-400 cm, and 400-500 cm soil layers on shady slopes in comparison to that on slope croplands at the 1-10 years, 11-20 years, 21-30 years, and ≥30-year stages. Water content in the 200-300 cm soil layer during the ≥30-year stage was smaller than that at the 1-10-year stage on sunny slopes ( Figure 4A). There were no dried soil layers on the sunny and shady slopes ( Figure 5). In addition, the water content of the sunny slopes was significantly lower than that of shady slopes in the 340-500 cm soil layer in the ≥30-year stage (p < 0.05).

Deep Soil Water at Different Stages on R. pseudoacacia Plantations
Over time, water content decreased significantly at the 11-20-year stage in the 200-300 cm, 300-400 cm, and 400-500 cm soil layers, and soil water content reached its lowest value, which was close or equal to the wilting coefficient, at the ≥30-year stage ( Figure 6). The available soil water storages were 0 and 8 mm in the 200-500 cm soil layer at the ≥30-year stage on sunny and shady R. pseudoacacia plantations, respectively. A dried soil layer started to appear at the 11-20-year stage, and it was widespread in the 200-500 cm soil layer (Figure 7).

Deep Soil Water at Different Stages on R. pseudoacacia Plantations
Over time, water content decreased significantly at the 11-20-year stage in the 200-300 cm, 300-400 cm, and 400-500 cm soil layers, and soil water content reached its lowest value, which was close or equal to the wilting coefficient, at the ≥30-year stage ( Figure 6). The available soil water storages were 0 and 8 mm in the 200-500 cm soil layer at the ≥30-year stage on sunny and shady R. pseudoacacia plantations, respectively. A dried soil layer started to appear at the 11-20-year stage, and it was widespread in the 200-500 cm soil layer (Figure 7).

Deep Soil Water at Different Stages on R. pseudoacacia Plantations
Over time, water content decreased significantly at the 11-20-year stage in the 200-300 cm, 300-400 cm, and 400-500 cm soil layers, and soil water content reached its lowest value, which was close or equal to the wilting coefficient, at the ≥30-year stage ( Figure 6). The available soil water storages were 0 and 8 mm in the 200-500 cm soil layer at the ≥30-year stage on sunny and shady R. pseudoacacia plantations, respectively. A dried soil layer started to appear at the 11-20-year stage, and it was widespread in the 200-500 cm soil layer (Figure 7).

Deep Soil Water at Different Stages on C. korshinskii Plantations
As shown in Figure 8, water content was 0.09, 0.08, 0.07, and 0.07 cm 3 cm −3 in the 200-500 cm soil layer at the 1-10 years, 11-20 years, 21-30 years, and ≥30-year stages, respectively; there were no significant differences among these values in the 200-500 cm soil layer, and so the variation is in the 200-300 cm, 300-400 cm, and 400-500 cm soil layers. Deep soil water content was significantly smaller than that on slope croplands. A dried soil layer extended through the 200-500 cm soil layer starting in the 1-10-year stage. The water content was close to the wilting coefficient (0.06 cm 3 cm −3 ) at the 11-20 years, 21-30 years, and ≥30-year stages (Figure 9), and the available soil water storages were 102, 47, 39, and 26 mm, respectively.

Deep Soil Water at Different Stages on C. korshinskii Plantations
As shown in Figure 8, water content was 0.09, 0.08, 0.07, and 0.07 cm 3 cm −3 in the 200-500 cm soil layer at the 1-10 years, 11-20 years, 21-30 years, and ≥30-year stages, respectively; there were no significant differences among these values in the 200-500 cm soil layer, and so the variation is in the 200-300 cm, 300-400 cm, and 400-500 cm soil layers. Deep soil water content was significantly smaller than that on slope croplands. A dried soil layer extended through the 200-500 cm soil layer starting in the 1-10-year stage. The water content was close to the wilting coefficient (0.06 cm 3 cm −3 ) at the 11-20 years, 21-30 years, and ≥30-year stages (Figure 9), and the available soil water storages were 102, 47, 39, and 26 mm, respectively.

Deep Soil Water at Different Stages on C. korshinskii Plantations
As shown in Figure 8, water content was 0.09, 0.08, 0.07, and 0.07 cm 3 cm −3 in the 200-500 cm soil layer at the 1-10 years, 11-20 years, 21-30 years, and ≥30-year stages, respectively; there were no significant differences among these values in the 200-500 cm soil layer, and so the variation is in the 200-300 cm, 300-400 cm, and 400-500 cm soil layers. Deep soil water content was significantly smaller than that on slope croplands. A dried soil layer extended through the 200-500 cm soil layer starting in the 1-10-year stage. The water content was close to the wilting coefficient (0.06 cm 3 cm −3 ) at the 11-20 years, 21-30 years, and ≥30-year stages (Figure 9), and the available soil water storages were 102, 47, 39, and 26 mm, respectively.

Response of Deep Soil Water to Vegetation Types and Slope Aspects
Vegetation restoration types significantly affected deep soil water in the study. In plant-soil-atmosphere systems, a plant acts as a pathway along which water transfers from the soil to the atmosphere, and its roots play a key role in the development of deep soil water [30]. The plant-soil-environment was in a relatively stable state on the natural abandoned slopes in the study because of nonsignificant differences in soil water content at all of the stages, indicating that low water consumption by the shallow root systems of the natural vegetation occurred in the deep soil layer in the study area. As the planted forests have higher evapotranspiration than precipitation recharge in the study region [7,31] and need to reach a larger water depletion depth [32], R. pseudoacacia plantations and C. korshinskii plantations had smaller soil water content in the deep soil layers in the study. This result was consistent with the results of similar studies [30,33]. As deep soil water cannot be replenished in time [17,26] when the deep soil water is exhausted, vegetation recovery of the R. pseudoacacia plantations and C. korshinskii plantations will be more restricted by climate conditions, especially in low flow years [15]. Thus, the recovery of natural vegetation was independent of deep soil water, and in contrast, deep soil water alleviated the need for artificial vegetation in some cases.
Generally, soil water conditions on shady slopes are better than those on sunny slopes in the hill and gully Loess Plateau region [34]. However, overall, there were no significant differences between the sunny and shady slopes on the natural abandoned slopes and R. pseudoacacia plantations in the study, indicating that the slope aspect was not one of the major factors influencing variations in soil water in the 200-500 cm soil layer in the hill and gully Loess Plateau region. In addition, we noticed another phenomenon; although there were no significant differences in soil water content in the 200-500 cm soil layers between the sunny and shady natural abandoned slopes, the soil water content of the sunny slopes was significantly lower than that of the shady slopes in the 340-500 cm soil layers at the ≥30-year stage. The B. ischaemum community and A. gmelinii community were the vegetation types at the ≥30-year stage on the sunny and shady natural abandoned slopes, respectively (Table 1). Relevant studies have shown that the depth of depleted soil water under a B. ischaemum community could reach 500 cm in the Loess Plateau, which is much deeper than that of the A. gmelinii community [26,35]. Therefore, the difference in soil water content between the sunny and shady slopes in the 340-500 cm soil layer might result from the difference in soil water depletion depths in B. ischaemum and A. gmelinii communities.

Sustainability of Deep Soil Water on Natural Abandoned Slopes
With vegetation recovery, deep soil water content did not change significantly, except in the 200-300 cm soil layer on sunny slopes, and there was no soil desiccation at the different stages. These results further showed that deep soil water was stable on the natural abandoned slopes, and natural vegetation was an ideal vegetation type for ecological restoration in the hill and gully Loess Plateau region. Furthermore, the main impact of vegetation recovery on deep soil water content was in the 200-300 cm soil layer, and despite this result, there was no dried soil layer in the soil layer. Yin et al. [26] and Chen et al. [36] noted that the infiltration depth was more than 300 cm under grasslands after storm rainfall with extremely high intensity and a long duration or in high flow years in the study area. Thus, soil water was easily replenished on natural abandoned slopes in the study area, and under the circumstance that soil desiccation did not occur, deep soil water was available for vegetation recovery.

Sustainability of Deep Soil Water on R. pseudoacacia Plantations
Over time, the soil water content decreased significantly on the R. pseudoacacia plantations in the study. This result was consistent with previous findings [23,37] and further showed that vegetation recovery exacerbated soil desiccation in the deep soil layer. The soil water content has been decreasing significantly since the 11-20-year stage, so it was close to or equal to the wilting coefficient at the ≥30-year stage. The dried soil layer began to be widely distributed in the 200-500 cm soil layer at the 11-20-year stage, showing that deep soil water was consumed, so that soil desiccation started to appear from the 11-20-year stage and resulted in soil desiccation. A deep soil layer would be unable to provide available water for vegetation recovery at the ≥30-year stage, further indicating that almost all of soil water that was used by vegetation was from the shallow soil layer (0-200 cm) in this stage. However, R. pseudoacacia plantations might not obtain enough soil water from the shallow soil layer in this stage because undergrowth vegetation had recovered obviously, while R. pseudoacacia began to degenerate (Table 1), and undergrowth vegetation might have a greater advantage to compete for soil water. A relevant study noted that the stable growth period of R. pseudoacacia is not long, unless soil water is sufficient in the study area [38]. Furthermore, vegetation is a vital and necessary part of ecosystem and its degeneration necessarily leads to series of reactions, such as the descent of soil quality. Thus, deep soil water could not be persistently provided, and soil desiccation limited the recovery of R. pseudoacacia plantations and then hindered sustainable development of R. pseudoacacia plantation in the study.
Given the situation that large tracts of R. pseudoacacia plantations are at the 10-20-year stage currently because of the "Grain for Green Project" that began in 1999, if the R. pseudoacacia plantations were managed in a timely manner to prevent further degradation of deep soil water as part of the "Grain for Green Project", the soil desiccation of R. pseudoacacia plantations would be controlled effectively. For the R. pseudoacacia plantation at the 21-30 years and ≥30-year stages, undergrowth vegetation covers a large area and can already control soil erosion to some degree [39]; therefore, reasonable thinning might be beneficial to relieve soil desiccation of the deep soil layer. Furthermore, suitable land preparation might be useful to improve deep soil water conditions. For example, a terrace is one of the main soil and water conservation strategies that can collect rainwater, increase water infiltration, and increase reserves of soil water [40]; thus, it might be constructed to remit or improve soil desiccation in the deep soil layer on the R. pseudoacacia plantations in the hill and gully Loess Plateau region.

Sustainability of Deep Soil Water on C. korshinskii Plantations
The water conditions of the C. korshinskii plantations were poor at every stage. Deep soil water was consumed rapidly during the 1-10-year stage in the study and was close to the wilting coefficient since the 11-20-year stage. Thus, for 21 years, there was no useable deep soil water for vegetation recovery.
As shown in Table 1, the coverage of C. korshinskii decreased since the 21-30-year stage. The table shows that the C. korshinskii plantations began to degenerate. In contrast, with the undergrowth vegetation recovering, the coverage of the undergrowth vegetation (a later succession community, Artemisia gmelinii community) [41] increased and reached 50% (Table 1), which enabled this vegetation to control soil erosion effectively in the study area [39]. A. gmelinii had an advantage in the competition of water absorption with C. korshinskii for the past 21-30 years. Considering the result that the A. gmelinii community did not result in water consumption in the deep soil layer, and with C. korshinskii plantations degenerating and the A. gmelinii community recovering in the later stages, soil desiccation would be ameliorated. Furthermore, appropriate manual intervention, such as cradling or clearing C. korshinskii plantations that have been in existence since 21-30 years ago or building terraces, could accelerate the recovery of soil water in the deep soil layer.

Conclusions
Natural vegetation restoration is an appropriate way to achieve sustainable ecosystems in consideration of deep soil water. In contrast, soil desiccation intensified until it was difficult for the vegetation to obtain available water in the deep soil layer on R. pseudoacacia or C. korshinskii plantations. Soil desiccation began to appear at the 11-20-year stage, and it became increasingly severe until the deep soil water was close to the wilting coefficient at the ≥30-year stage on R. pseudoacacia plantations. Deep soil water was rapidly consumed, and soil desiccation began to appear during the 1-10-year stage, and then deep soil water was close to the wilting coefficient at the later stages on C. korshinskii plantations. According to the results, the plantations needed to be managed in a timely manner to prevent or relieve soil desiccation.