Fish-Scale Pits with Diversion Holes Enhance Water Conservation in Semi-arid Loess Soil: Experiments with Soil Columns, Mulching, and Simulated Rainfall

Fish-scale pits, as one of the most important ecological engineering in semi-arid areas, especially in the Loess Plateau, China, effectively reduced water and soil loss by intercepting runoff. However, the water saved by the pits is mainly accumulated in the upper soil layer and thus easily loosed by ineffective evaporation. In this study, soil column simulation experiments were conducted to observe the effects of diversion holes on soil evaporation and soil water distribution in the loessial soil under two rainfall depths and different soil surface treatments. The soil samples were passed through a 2-mm sieve. The soil filling height and the soil volume was 140 cm and 54,652.17 cm3 in each soil column, respectively. For each treatment, there were three replicates. The results showed that gravel-mulching (GM), straw-mulching (SM), and diversion hole (DH) significantly decreased soil water evaporation compared with non-mulching (CK) under the different rainfall depths (P < 0.05). Under rainfall of 100 mm, the DH, GM, and SM caused an increase in the soil water content (SWC) in the 80–120 cm layer by 101.0%, 23.6%, and 65.5%, respectively. Under rainfall of 30 mm, CK increased SWC in the 0–40 cm layer, both GM and SM increased the SWC in the 0–60 cm layer, and the DH treatment allowed an increase in the 0–120 cm layer. Our results suggest that DH was a better measure to reduce soil evaporation and improve rainfall resource utilization, especially under high rainfall events. Our research provides scientific basis for the technology of soil deep spatiotemporal control and is of great significance to the balance and restoration of the ecological environment in the semi-arid Loess Plateau. Graphical Abstract Diversion hole has induced responses on soil water evaporation and distribution under two rainfall intensity Diversion hole has induced responses on soil water evaporation and distribution under two rainfall intensity


Introduction
The Loess Plateau is a typical arid and semi-arid region with loess thickness of 0-350 m , where the groundwater level is low. Rainfall is almost the only source of soil water supply and most of them are heavy hard rain. The average annual precipitation is 300-500 mm with approximately 60-70% falling from June to September (Li et al. 2012;Gao et al. 2014). Runoff and soil erosion occur after heavy rainfall because rainfall intensity exceeds infiltration capacity (Fang et al. 2008;Angulo-Martínez and Beguería 2009;Zhao et al. 2013;Wei et al. 2014;dos Santos et al. 2017;Chen et al. 2018). The coexistence of drought, water shortage, and soil erosion is the key factor to restrict the sustainable development of dry farming and is also the fundamental cause of ecological fragility in the Loess Plateau (Ruiz-Sánchez et al. 2005;Song et al. 2017;Shi et al. 2018).
A series of soil and water conservation engineering measures have been popularized to solve these problems, and fishscale pit is one of the most effective measures among them in semi-arid areas of Loess Plateau Zhang et al. 2014;Yu et al. 2017). Fish-scale pit could effectively reduce Highlights • Fish-scale pits with diversion holes can transport rapidly surface water into the deep soil layers for soil and water conservation.
• Diversion hole can reduce soil water evaporation and increase soil water content in the deep soil layer.
• Diversion hole is a better measure to improve rainfall resource utilization, especially under high rainfall events. * Yu Liu kingliuyu@nwafu.edu.cn 1 the surface runoff during heavy hard rain by intercepting rainwater on the soil surface to the pit and increasing water infiltration (Fu et al. 2010;Feng et al. 2018). However, fish-scale pit also increased the ineffective evaporation because of the large contact area between soil and air . Li et al. (2018) found that the fish-scale pit without mulching experienced in jujube orchard significantly increased evaporation by 42.28% and 65.59%, respectively, during the two growing seasons. Moreover, most of the fish-scale pits would be filled with sediment or destroyed by runoff in a few years (Wang et al. 2011). Meanwhile, Zhang et al. (2016a, b) found that the evaporation has increased considerably over the past 50 years by analyzing meteorological observation data from 1960 to 2010 in Dingxi, a typical semi-arid region of the Loess Plateau. Besides, Li et al. (2012) reported that the upward trend of soil potential evapotranspiration in the Loess Plateau will be more pronounced after 2050. Therefore, the key to solve the problem of soil and water erosion in the Loess Plateau region is to prevent invalid water evaporation, control rainfall runoff and realize rainfall resource utilization (Hu et al. 2014).
Adding mulch on soil surface is a common way to effectively reduce water loss (Chiroma et al. 2006;Wang et al. 2009Wang et al. , 2016Huo et al. 2017;Fu et al. 2018). It can weaken the impact of splash erosion of raindrops on soil surface, protect soil from erosion caused by rainfall, and reduce runoff (Rees et al. 2002;Findeling et al. 2003;Gholami et al. 2013). Moreover, effectively reducing evaporation (Qin et al. 2006;Mahdavi et al. 2017) and improving soil infiltration capacity (Wang et al. 2016;Jordán et al. 2010) owing to the physical barriers, which was set on the soil surface, hindered and slowed the exchange of water and energy between soil surface and atmosphere. Li et al. (2016) found that compared with single fish-scale pits, the fish-scale pit with mulching significantly increased soil water storage and water compensation in jujube orchard owing to the increase of water infiltration and the decrease of evaporation. Gravel and straw are widely used as cover materials in the Loess Plateau (Xie et al. 2010;Ma and Li 2011;Fan et al. 2014;Wang et al. 2015).
The Loess Plateau is known as Bnature soil reservoir,^because the deep soil layer has huge storage capacity for water (Zhu 2006). However, factors such as less rainfall, more runoff, high evaporation intensity, and excessive soil water depletion by vegetation of the deep soil moisture may result in a dried soil layer (Jipp et al. 1998;Chen et al. 2008;Wang et al. 2018). Sun et al. (2018) found that long-term cultivation of alfalfa pastures seriously consumed the deep soil water. This process of soil water deficit of alfalfa pastures was obviously aggravated after 4 years of planting. Therefore, another way for improving rainfall efficient utilization should concentrate on reducing runoff on the slope, and quickly transporting and storing the surface water into deep soil, thereby reducing the evaporation loss of surface layer, giving full play to the role of Bsoil reservoir.P revious studies found that the mixture of the irrigated water and fertilizer flows into the pits in the management of apple orchard in northern China, which is useful for absorbed of apple roots (Zhao et al. 2014a, b). However, fewer researches have discussed the way of rapidly transporting surface water into the deep soil layers for soil and water conservation, whether it can increase the water content in the deep soil layers and reduce the ineffective evaporation is still a scientific problem to be solved. In this context, soil column simulation experiments were conducted. The major objectives and purposes of this research were to (a) determine the effects of mulch and diversion hole measures on soil evaporation under different rainfall intensity, (b) analyze the effects of mulch and diversion hole measures on the water distribution in a loessial soil under different rainfall intensity, and (c) try to find the more suitable water conservation measure for rainfall efficient utilization and the technology of soil deep spatiotemporal control in the Loess Plateau.

Soil and Experimental Equipment
The soil used in this study is the loessial soil and was collected from Nangou Village, Ansai County, Yan'an City, Shaanxi Province, China (109°19′ E, 36°51′ N). The experiment was carried out in the Institute of Soil and Water Conservation in Yang Ling District, Shaanxi Province, China. The soil samples were air-dried at room temperature, crushed manually, and then passed through a 2-mm sieve. The soil collected from field was high on soil spatial heterogeneity. Difference in soil pore distribution was obvious owing to the coarse fragments which could cause preferential flow in the process of water infiltration and redistribution. We sieved the soil and removed the coarse fragments for the homogeneity of soil. The particle composition of the treated soil was analyzed by laser particle size analyzer; the soil basic physical properties are shown in Table 1. The soil columns consist of column bodies that were made of PE (polyethylene) material, and plugs were made of PVC (polyvinyl chloride) (Fig. 1). The size of the column was 150 cm in height and 22.3 cm in internal diameter, the soil filling height was 140 cm, and the total volume of the soil in each soil column was 54,652.17 cm 3 . White gauzes were placed at the bottom of the column before the soil columns were packed with treated soil, and Vaseline was uniformly daubed on the inner wall of the soil column to eliminate the effect of wall dominant flow. The soil bulk density in columns was 1.3 g cm −3 after packing, the soil filling way was layered filling with a thickness of 5 cm each time, and it is necessary to fluffed in the end of each layer to avoid the changes of soil structural and hydrodynamic between layers. The measuring tubes of TDR (TRIME-PICO IPH/T3, IMKO, Germany) were installed during the soil packing process.

Experiment Design
Four treatments were set in this experiment, including non-mulching (CK), gravel-mulching (GM), strawmulching (SM), and diversion hole (DH). For each treatment, there were three replicates. Gravel (1-2.5 cm in diameter, 1.34 kg weight, and 1.5 cm thick in final) and maize straw (5-7 cm in length, 0.07 kg weight and 2.7 cm thick in final) were used to cover the soil in the GM and SM treatments, respectively. PVC tubes (3 cm in diameter, 80 cm in length) with holes in its inner wall were placed in the center of the soil column in the DH treatment.
The size of each hole was 3 mm in diameter in the wall of PVC tube. The holes were set every 2 cm in the PVC tubes, where there were two holes in each cross section. The total number of holes was 78 in each PVC tube. There were two sprinkler processes that simulated natural rainfall in this experiment. The first sprinkler amount was 3.96 kg water (equal to rainfall of 100 mm) and carried out on May 14, 2017 with 65 days (May 15, 2017-July 18, 2017) of evaporation. The second was 1.12 kg water (equal to rainfall of 30 mm) and carried out on September 21, 2017 with 25 days (September 22, 2017-October 17, 2017) of evaporation. Soil evaporation of each column was measured by using the weighing method, and the soil water content was measured by TDR every 3-5 days ( Fig. 1) up to a measurement depth of 120 cm every

Statistical Analysis
According to the soil water balance method, the cumulative evaporation (Ec) was calculated as follows: where Wo (g) is the initial weight of the soil column in the beginning of the experiment. Wi (g) is the weight of the soil column on the ith day. ρ (g cm −3 ) is the density of water. S (cm 2 ) is the evaporation area of the soil column. According to Ec and days of evaporation, the daily evaporation (Ed) was calculated as follows: where Ec i (mm) is Ec on the ith day, Ec i−d (mm) is Ec on the (I − d)th day, and d is the evaporation days. The comparison of the results between the different treatments would be influenced by the different initial water supply (h, in mm). Therefore, we used the ratio of evaporative loss (the proportion of the final Ec to the initial water supply of the soil in the evaporation time (Li and Li 1991)) as a more appropriate index to illustrate evaporative effects. If the ratio is high, the treatment is relatively prone to water evaporation. The evaporative loss ratio (R) was calculated as follows: Statistical analysis was performed with SPSS 20.0. Oneway analysis of variance (ANOVA) was conducted to test the effects of each treatment on soil moisture and evaporation. Significant differences were detected using the leastsignificant difference (LSD) test at P < 0.05. The plotting was made with Origin 8.5.

Soil Daily Evaporation
The average Ed in different measures under two rainfall intensity is shown in Fig. 2. The average Ed of CK, DH, SM, and GM was 1.48 mm, 1.31 mm, 1.17 mm, and 0.96 mm under rainfall of 100 mm, respectively, and was 0.88 mm, 0.58 mm, 0.32 mm, and 0.26 mm under rainfall of 30 mm, respectively. The average Ed of different treatments under both rainfall intensity were CK > DH > SM > GM, and significant differences were found between the different treatments (P < 0.05). The changes of Ed in different time under two rainfall intensity are shown in Fig. 3. After rainfall of 100 mm, the Ed of CK and DH, GM, and SM had the same trend, and the Ed of DH and GM was lower than that of CK and SM, respectively. The Ed of both CK and DH in 1-22 days was higher than in 23-65 days because the soil water decreased with time. The average Ed was 2.70 (1.25 to 3.99) mm in 1-22 days and 0.82 (0.35 to 1.14) mm in 23-65 days under CK, 2.28 (0.73 to 3.59) mm in 1-22 days, and 0.79 (0.48 to 1.13) mm in 23-65 days under DH. The average Ed was 0.96 (0.62 to 1.33) mm under GM, and 1.17 (0.72 to 1.60) mm under SM (Fig. 3a). After rainfall of 30 mm, the average Ed of CK, DH, SM, and GM were 0.88 mm, 0.58 mm, 0.32 mm, and 0.26 mm, respectively. The Ed decreased with the increase of evaporation days with a larger reduction rate in the first 7 days. The difference test showed that there were significant differences (P < 0.05) between DH and CK except for evaporation in 0-3 days and 20-25 days, and compared with CK, both the GM and SM significantly decreased the evaporation during the whole evaporation process (P < 0.05) (Fig. 3b).

Soil Cumulative Evaporation
The changes of Ec in different measures under two rainfall intensity are shown in Fig. 4. After rainfall of 100 mm, the Ec of CK, DH, SM, and GM in 65 days were 96.46 mm, 85.43 mm, 75.86 mm, and 62.39 mm, respectively. Compared with CK, the GM and SM decreased evaporation Standard errors of the means (n = 3) are presented as ± values GM gravel-mulching, SM straw-mulching, DH diversion hole, CK non-mulching by 35.32% and 21.36% (P < 0.05), respectively, and the DH decreased evaporation by 11.44% (P < 0.05). However, no significant effects were observed in 0-9 days, and there was significant difference among different treatments in 12-65 days (P < 0.05) (Fig. 4a). After rainfall of 30 mm, the Ec of CK, DH, SM, and GM in 0-25 days were 21.95 mm, 14.59 mm, 8.03 mm, and 6.59 mm, respectively. Compared to CK, the GM and SM significantly decreased evaporation by 69.97% and 63.44% (P < 0.05), respectively, the DH decreased evaporation by 33.55% (P < 0.05), whereas no significant difference in 0-3 days, and compared with CK, all the SM, GM, and DH significantly decreased the evaporation in 7-25 days (P < 0.05) (Fig. 4b).

Soil Evaporative Loss Ratio
Compared with CK, the DH, SM, and GM decreased the soil evaporative loss ratio by 12.82%, 25.08%, and 38.83% after rainfall of 100 mm, respectively (P < 0.05), and decreased by 38.61%, 66.93%, and 76.19% after rainfall of 30 mm, respectively(P < 0.05). Therefore, all the DH, SM, and GM could reduce soil evaporation in various degree (Table 3).

Rainfall of 100 mm
The treatments GM, SM, and DH had different effects on the soil water content in the different soil layers under the rainfall depth of 100 mm (Fig. 5). Rainfall in the first day, both the maximum value of soil water and its increment under CK, GM, SM, and DH appeared at 40 cm, 40 cm, 50 cm, and 90 cm layers, respectively, where the maximum value of soil water content was 37.62%, 39.96%, 38.29%, and 31.70%, respectively, and increased by 11.74%, 14.33%, 13.07%, and 11.26%, respectively. All the GM, SM, and DH significantly increased the soil water content in deep soil layer; compared with CK, the GM and SM increased soil water content in 40-80 cm layer by 9.55% and 12.69%, respectively, GM, SM, and DH increased the water content in 80-120 cm layer by 23.59%, 65.45%, and 101.00%, respectively. With infiltrating of the soil water, the change trend of soil water distribution under CK, GM, SM, and DH was basically the same starting from the second day. The water distribution of soil Changes of the daily evaporation under rainfall of 30 mm. Date were presented as the means (n = 3) ± SD. All raw data were compared using single-factor variance analysis method with the least-significant difference (LSD) test at P < 0.05. CK non-mulching, GM gravel-mulching, SM straw-mulching, DH diversion hole Fig. 2 The average daily evaporation of different treatments under two rainfall intensity. Date were presented as the means (n = 3) ± SD. All raw data were compared using single-factor variance analysis method. Different letters indicate significant difference (LSD) among the treatments at P < 0.05. CK non-mulching, GM gravel-mulching, SM strawmulching, DH diversion hole profile tended to be stable under each treatment starting from the tenth day. Compared with CK, all the GM, SM, and DH increased soil water content in 0-80 cm layer, the GM decreased soil water content in 100-120 cm layer, and both the SM and DH decreased soil water content in 80-120 cm layer. The soil water content in each layer under both GM and SM was higher than CK starting from the 19th day, and the DH decreased water content in 0-80 cm layer, compared with CK, whereas increased water content in 80-120 cm. Briefly, the DH increased the soil water in deep layer.

Rainfall of 30 mm
The spatiotemporal changes of soil water distribution in the different measures under rainfall of 30 mm are shown in Fig. 6. The rainfall of 30 mm increased the water content in 0-40 cm layer under CK, increased the water content in 0-60 cm layer under both GM and SM, and increased the water content in 0-120 cm layer under DH. The maximum value of soil moisture increment appeared at 20 cm layer under all the CK, GM, SM, and DH, where the soil water content increased by 10.44%, 9.63%, 5.59%, and 3.19%, respectively. The changes of soil water under CK were mainly at the 0-40 cm soil layer. Both GM and SM increased the water content in 0-60 cm layer with evaporation time. DH increased soil water content in 0-120 cm layer.

Discussion
Soil evaporation potential depends on soil hydraulic conductivity and atmospheric evaporation potential (Gardner and Fireman 1958). In this experiment, the results showed that all the GM, SM, and DH could restrain soil water evaporation. Both the GM and SM could avoid direct solar radiation to consequently decrease atmospheric evaporation potential by setting physical barrier layers on the soil surface (Mahdavi et al. 2017). Based on energy balance, SM could also appreciably decrease soil evaporation by reducing the latent heat flux . GM could reduce the actual evaporation because the gravel could block the channel of water circulation by blocking the soil pores and preventing capillary water from rising (Li 2003;Zhang et al. 2016a, b), and the greater thickness of gravel, the greater its inhibition (Modaihsh et al. 1985). DH effectively decreased the water evaporation on soil surface because it could directly transport and storage water in Fig. 4 a Changes of the cumulative evaporation under rainfall of 100 mm. b Changes of the cumulative evaporation under rainfall of 30 mm. Date were presented as the means (n = 3). All raw data were compared using single-factor variance analysis method with the leastsignificant difference (LSD) test at P < 0.05. CK non-mulching, GM gravel-mulching, SM straw-mulching, DH diversion hole Standard errors of the means (n = 3) are presented as ± values. All raw data were compared using single-factor variance analysis method. Different letters indicate significant difference (LSD) among the treatments at P < 0.05 CK non-mulching, GM gravel-mulching, SM straw-mulching, DH diversion hole deep soil layer. In this study, the effects of mulching measures on retraining soil water evaporation were better than DH. Because DH had the large contact area between soil and air to increase the ineffective evaporation, the same as CK. Compared with CK, both the GM and SM had no significantly effects on Ed because the atmospheric evaporation potential was significantly weakened by GM and SM. However, the Ed under both CK and DH were higher at the early stage (1-22 days under rainfall of 100 mm and 1-7 days under rainfall of 30 mm) and lower at the latter stage (23-65 days under rainfall of 100 mm and 8-25 days under rainfall of 30 mm). The soil evaporation intensity at early stage was higher, because high soil water content increased the soil hydraulic conductivity. The soil water in the lower layer moved continuously upward along the capillary pore due to the action of matrix suction, it can supplement the soil shallow layer water loss caused by evaporation. The soil evaporation intensity at latter stage was lower, because soil water content reduced by evaporation, soil hydraulic conductivity was decreased, the continuous water flow in soil pores was gradually weakened, and the replenishment of soil shallow layer water became difficult, which was insufficient to meet the evaporation requirements of the surface environment (Zribi et al. 2015). The texture of the loessial soil is loam with large number of medium and small soil pores, which would help water infiltration, whereas it would be bad for soil water holding capacity (Wu et al. 2015;Dong et al. 2018). Soil water conductivity increased with the increase of soil water potential, which eventually led to an increase in the soil water infiltration rate (Stewart et al. 2013;Luo et al. 2018). Due to the difference of water content between upper and lower soil layers, the rainwater would infiltrate continuously with the gradients of Fig. 5 Temporal and spatial changes of soil water content in each treatment under rainfall of 100 mm. Date were presented as the means (n = 3). CK nonmulching, GM gravel-mulching, SM straw-mulching, DH diversion hole gravity potential and matrix potential in the soil profile. In this study, under rainfall of 100 mm, all the GM, SM, and DH significantly increased the soil water content in deep soil layer, compared with CK, the GM, SM, and DH increased the water content in 80-120 cm layer by 23.59%, 65.45%, and 101.00%, respectively. Under rainfall of 30 mm, the changes of soil water under CK were mainly observed at the 0-40 cm soil layer. Both GM and SM favored an increase of the soil water content in the 0-60 cm layer with evaporation time, and DH increased the soil water content in the 0-120 cm layer. Under the heavy rainfall, water can infiltrate into the deep soil layer and increase the soil water content in different treatment. DH was more easily transport and storage rainwater to soil deep layer, because the water would infiltrate into the hole wall, whereas under the small rainfall, SM and GM cannot increase the water content in the deep soil layer under evaporation, except for DH. The water infiltration time after rainfall under natural conditions was affected by the factors including rainfall, rainfall intensity, the properties of unsaturated zone, and soil initial water content (Rose 1996;Chiu et al. 2009;Liu et al. 2011;Ojha et al. 2017), and the evaporation had no significantly effects on the redistribution of water infiltration under a large amount of rainfall. Liu et al. (2015) found that with the increase of rainfall, there was a relatively small change of the evaporation when the rainfall is greater than 20 mm. In general, the effects of mulching measures (especially Fig. 6 Temporal and spatial changes of soil water content in each treatment under rainfall of 30 mm. Date were presented as the means (n = 3). CK nonmulching, GM gravel-mulching, SM straw-mulching, DH diversion hole gravel-mulching) on decreasing evaporation were better than DH. Hence, DH with mulching would be the best water conservation measure, which not only directly transported and stored rainwater into the deep soil but also restrained the evaporation.

Conclusions
In general, the effects of diversion hole on limiting soil water evaporation were better than those observed non-mulching, favoring an increase of the deep soil water content. Under the high rainfall, diversion hole was better than mulching measures for increasing the water content in the deep soil layer. Our results showed that diversion hole not only reduced soil evaporation but also was a better measure to decrease runoff loss on soil surface and improve rainfall resource utilization. We suggest that diversion hole with mulch should be closely considered for use in the soil and water conservation in the Loess Plateau region.

Compliance with Ethical Standards
Conflict of Interest The authors declare that they have no conflict of interest.