The study area isDajigouCatchment(Table1; Fig.1), a typical loess hilly areain thenorthwestern part of theChinese Loess Plateau withinWuqi County of Shaanxi Province (36° 33′ 33″-37° 24′ 27″N, 107° 38′ 37″-l08° 32′ 49″E; 1233-1809 ma.s.l.). The mean annual precipitation was about 464.5 mm (1957-2013), with 61% falling in the rainy season from July to September. The monthly temperature ranged from -28.5° C(December, 1967) to 38.3° C(July, 2001), with an annual mean temperature of 7.9° C(1957-2013). The surface soils in the study area are of loessial origin with relatively coarse particles(Chronicle, 1991), and the natural vegetation (presumably steppe-dominated) has almost completely disappeared due to a prolonged human disturbance.To restore theseverely disturbed ecological environment, Chinese government has recently implemented the Grainfor Green Program through vegetationrestoration measures.

Fig. 1

	Figure Option ViewDownloadNew Window
	Fig. 1 Location of the study area and distribution of the five plots within Dajigou Catchment.RPa, Hippophaerhamnoides+Pinustabuliformis; RPb, H.rhamnoides+P.tabuliformis; P, P.tabuliformis; R, H.rhamnoides; G, Lespedeza davurica+Leymussecalinus.

Taking into consideration of thetopographical features and the vegetation types, five plots (20 m× 5 m) were constructed within the study area in July 2009. The vegetation types in the five plots are:Hippophaerhamnoides+Pinustabuliformis (Plot RPa), H.rhamnoides+P.tabuliformis (Plot RPb), P.tabuliformis(Plot P), H.rhamnoides (Plot R), and Lespedeza davurica+Leymussecalinus(Plot G) (Table 1; Fig. 1). The vegetation and surface soil were severely disturbedat the five plots through trampling and digging during the construction and we termed the period immediately after the construction the early stage after land disturbance (ESLD). The vegetation and soil conditions at the five plots have graduallyameliorated over time and we termed the period of amelioration the restoring stage after land disturbance (RSLD). Figure 2 shows the condition contrasts between ESLD and RSLD for the Plot G (Figs. 2a and c) and for the Plot P (Figs. 2b and d).

Table 1

Table 1

Table 1 Specific conditions of five runoff plots Item RPa RPb P R G Vegetation type H.rhamnoides +P.tabuliformis H.rhamnoides +P.tabuliformis P.tabuliformis H.rhamnoides Lespedeza davurica+ Leymussecalinus Density (plants/hm2) Pinus tabuliformis 800 700 1200 2300 - Hippophaerhamnoides 1500 1400 3000 2300 - Height (m) 1.88 1.98 3.33 2.62 - Slope gradient (° ) 12 29 17 17 28 Slope aspect ES37° ES35° WS12° NE34° WS3° Elevation (m) 1396 1380 1386 1406 1398 Canopy density (%) ESLD 48 32 40 35 55 RSLD 63 50 53 55 81 Note:ESLD, early stage after land disturbance; RSLD, restoring stage after land disturbance.

Table 1 Specific conditions of five runoff plots

Fig. 2

	Figure Option ViewDownloadNew Window
	Fig. 2 Condition contrasts between early stage after land disturbance (ESLD) and restoring stage after land disturbance (RSLD) for the Plot G (a and c) and for the Plot P (b and d)

At all of the five plots, we monitored the runoff production and sediment yield of each rainfall event during the rainy season (July-September) from 2009 to 2012. At the lower end of each plot, a sump was installed to collect sediment and runoff and the sump was made of concrete with a dimension of 1 m× 1 m× 1 m (Fig. 3a). A simple automatic meteorological field station (HOBO weather station, Onset Computer Co., Boerne, MA, USA), including a tilting rain gauge, was installed to record year-round meteorological data (Fig. 3b) and the data recording interval was5 min. The data thus allowed us to calculate the I₅, I₁₀, I₁₅ and I₃₀ values (note: I₅, I₁₀, I₁₅ and I₃₀were the maximum rainfall intensities for 5-min, 10-min, 15-min and 30-min, respectively).

Fig. 3

	Figure Option ViewDownloadNew Window
	Fig. 3 Sedimentand runoff collecting sump (a) and automatic meteorological station (b)

Three soil pits were excavated at the uphill, middle and downhill sitesand soil samples were collected from each pit at the depths of 0-20, 20-40, 40-60, 60-80 and 80-100 cm. Soil bulk densities were tested using a ring knife with a diameter of 5 cm and a height of 5 cm. Soil infiltration apparatuswas employed for recording the process of water infiltration into the soil and the depth of infiltration was calculated by an empirical Equation 1.

H=0.19635× h× cosa. (1)

Where H is the depth of infiltration; h is the change in the standing water level; and ais the slope gradient. At the beginning of the experiment, data were recorded every 10 s for 90 s; then, data were recorded every 30 s for 5 min; at the end, data were recorded once one minute. The experiment were repeated for 5-6 times toinsure data consistency (Zhao et al., 2010).

We used the Principal Coordinates Analysis method to convert the qualitative variables, such as vegetation and slope aspect, into quantitative variables (Zhang, 2004) to obtain the characteristic value of each variable. And, we then used the characteristic value for further analysis and the further analysis was carried out using the grey correlation method(Deng, 2002). This grey correlationmethod identifies the similarity or dissimilarity of the development trends among factors. By comparing a sequence of an established family of curveswith a reference sequence curveusing geometric similarity, one can determine the degree of connection between the reference sequence and thecomparingsequence. It means that the more similar, the greater degree the similarity is. The comparing sequence and the reference sequence can be both temporal series and non-temporal series. The method is shown as below:

Where are the new values obtained when the parameters are standardized by Equation 2, and x_i(k) are the original parameters.

The correlation coefficient is calculated by Equations 3-6.

Where Δ x_0i(k)is the absolute value of the difference between the comparing sequence and the reference sequence and ξ is the distinguishing coefficient. The value of ξ ranges from 0 to 1, but ξ =0.5 is considered to be a threshold for showing similarity. the correlation coefficient.

The grey relational grade (GRG, Γ ) is calculated by Equation 7.

We selected runoff production and sediment yield as the reference sequences and used several indicators as our comparing sequences. The used comparing sequences include vegetation type, vegetation coverage, rainfall amount, rainfall duration, average rainfall intensity, I₅, I₁₀, I₁₅, I₃₀, soil bulk density, soil infiltration rate, and slope aspect, and gradient. Then, the GRG was calculated for the references and the comparing sequences (Tables 2 and 3). According to Deng (2002), a larger GRGmeans a closer relationship between the comparing sequence and the reference sequences.

Fig. 4

	Figure Option ViewDownloadNew Window
	Fig. 4 Runoffproduction and sediment yield during ESLD and during RSLD for five different vegetation types

As shown in Table 2, the slope aspect had the strongest impact on the runoff production with a GRG of 0.6681 during ESLD. The second rankwas the soil infiltration rate with a GRG of 0.6524. The rainfall amount ranked as the third with a GRG of 0.6417, followed by rainfall duration with a GRG of 0.6303. Among the rainfall intensities, I₁₅had the largest GRG (0.6209). The influences of vegetation type and cover on the runoff were at the intermediate ranks among the 13 considered factors.It should be noted that the smallest GRG value of0.5406 (I₃₀) was still larger than the presumed threshold of 0.5, meaning that all the 13 considered factors were closely related to the runoff production.

During RSLD (Table 2), the rainfall duration with a GRG of 0.7443replaced the slope aspect as the most critical factor affecting the runoff production. The vegetation type ranked the second with a GRG of 0.6757 and the rainfall amount ranked the third with a GRG of 0.6415. The GRG of the soil steady infiltration rate ranked the fourth(GRG=0.6231). Among the rainfall intensities, I₁₀ was the most critical factor with a GRG of 0.6186. The influence of slope aspect on the runoff production during RSLD (eleventh rank) was much less important than during ESLD (first rank).

Table 2

Table 2

Table 2 Grey relational grade between runoff and its influential factors ESLD RSLD GRG Rank Proportion(%) GRG Rank Proportion (%) Vegetation Vegetationtype 0.6133 6 24.76 0.6757 2 25.94 Vegetation coverage 0.6023 7 0.6061 8 Rainfall Rainfall amount 0.6417 3 24.11 0.6415 3 25.39 Rainfall duration 0.6303 4 0.7443 1 Average rainfall intensity 0.5835 10 0.5675 12 I5 0.5584 12 0.6146 7 I10 0.5662 11 0.6186 5 I15 0.6209 5 0.6032 9 I30 0.5406 13 0.6012 10 Soil Soil bulk density 0.5936 9 25.39 0.5665 13 24.08 Soil steady infiltration 0.6524 2 0.6231 4 Topography Slope aspect 0.6681 1 25.74 0.5995 11 24.59 Slope gradient 0.5955 8 0.6156 6 Note:I5, 5-min maximum rainfall intensity; I10, 10-min maximum rainfall intensity; I15, 15-min maximum rainfall intensity; I30, 30-min maximum rainfall intensity; GRG, grey relational grade.

Table 2 Grey relational grade between runoff and its influential factors

Table 3 shows thatthe soil bulk density was the dominant factor affecting sediment yield during ESLD and the GRG was as high as 0.8113. The average rainfall intensity ranked the second with a GRG of 0.7444, followed by the soil steady infiltration rate (GRG=0.7128). Among the rainfall intensities, I₃₀ had the closest relationship with sediment yield (GRG=0.6890). The GRG was 0.6487for vegetation type and 0.6229 for vegetation coverage. The smallest GRG value of 0.5945 (slope gradient) was larger than the presumed threshold of 0.5, indicating that all of the considered 14 factors had close relationships with the sediment yield (note: runoff was added as an additional factor to the 13 factors listed in Table 2).

During RSLD, the four-grade rainfall intensities were all among the most significant factors, e.g., I₅ranked the second (GRG=0.7953), I₁₀the first (GRG=0.8012), I₁₅ the third (GRG=0.7779), and I₃₀the fourth (GRG=0.7593), indicating that rainfall intensity was the most important in influencing the sediment yield during RSLD. Similarly, the rainfall duration was promoted from rank tenth during ESLD to rank fifth during ESLD. By comparison, the soil bulk density was demotedfrom rank first during ESLD to rank tenth during RSLD. At the same time, the importance ofthe soil steady infiltration rate was demoted from rank third during ESLD to rank eighth during RSLD and the importance of runoff was promoted from rank thirteenth during ESLD to rank eleventh during RSLD. The importance of both the vegetation type and vegetation coverage on sediment yield was lowered during RSLD compared with that during ESLD.

Significant differences were observed under the different vegetation types in terms of runoff production and sediment yieldbetweenESLD and RSLD (Fig. 4). The runoff production and sediment yield during ESLD were remarkably higher than those during RSLD. For example, the runoff production was 13.55-foldhigher duringESLD than during RSLD and the sediment yield was 3.13-fold higher duringESLD than during RSLD under forest conditions (e.g., Plot P). The runoff production and sediment yield were 8.59-fold and 1.70-fold higher during ESLD than during RDLS, respectively, under grassland conditions (e.g., Plot G). Under mixed vegetation conditions (e.g., PlotsRPa, RPb, and R), the runoff production was 3~5 times higher during ESLD than during RSLD and the sediment yield was 1~2 times higher during ESLD than during RSLD. The analytical results demonstrate that vegetation restoration can effectively reduce the runoff production and sediment yield.

Table 3

Table 3

Table 3 Grey relational grade between sediment yield and its influential factors ESLD RSLD GRG Rank Proportion(%) GRG Rank Proportion(%) Runoff Runoff 0.6129 13 18.62 0.6323 11 19.32 Vegetation Vegetation type 0.6487 7 19.31 0.6119 13 18.73 Vegetation coverage 0.6229 11 0.6141 12 Rainfall Rainfall amount 0.6198 12 20.22 0.6627 9 22.54 Rainfall duration 0.6316 10 0.6926 5 Average rainfall intensity 0.7444 2 0.6758 6 I5 0.6668 5 0.7953 2 I10 0.6595 6 0.8012 1 I15 0.6482 8 0.7779 3 I30 0.6890 4 0.7593 4 Soil Soil bulk density 0.8113 1 23.15 0.6532 10 20.14 Soil steady infiltration rate 0.7128 3 0.6652 8 Topography Slope aspect 0.6362 9 18.69 0.6658 7 19.26 Slope gradient 0.5945 14 0.5949 14

Table 3 Grey relational grade between sediment yield and its influential factors

The importanceof different factors on runoff production differed significantly during ESLD than during RSLD. The importance of the major factors had a descending order during ESLD as follow: topography> soil> vegetation> rainfall, andthe descending order during RSLD became: vegetation> rainfall> topography> soil (see Table 2).In this study, the surface soil was severely disturbed and the vegetation was greatly reduced during the plot construction.When the surface soil was disturbed or/and compacted and the vegetation coveragewas reduced or removed, the soil steady infiltration rate and the canopy interception capacity were reduced, leading to increase in runoff production(Wilcox et al., 2003; Zhang et al., 2009; Wei et al., 2014). It should be stressed that vegetation root systems can improve the soil infiltration capacity and that the soil disturbance-resulted destruction of the root systems can increase runoff production through decreasing the soil steady infiltration rate (Gyssels et al., 2005; Wang et al., 2013, 2014a, b). Another point needs addressing here. That is, the importance of vegetation types on runoff production was relatively weak during ESLD (see Table 2) and the speculated reason is that the interception and retention capacities were rather limited at all five plots during ESLD simply due to rather limited vegetation coverage immediately after the plot construction. Consequently, rainfallbecame more important (the fourth rank) simply due to the reduced interception and retention capacities(Ai et al., 2013).

Researchers have extensively explored the effects of land disturbance on runoff production and sediment yield and the published literatures have provided more details that are quite relevant to our study. For example, Wilcox et al. (2003)and Vanacker et al. (2014)noted that surface soil disturbances can modify surface topographical features and the vegetation patch structure, eventually decreasing water storage or retention capacity on the hillslope. Mohr et al. (2013) found that the impact of micro-topography on surface runoff connectivity and water-repellent properties is the first-order control on hydrological and soil erosion processes. Similarly, Researchers have also extensively explored the effects of land restoration on runoff production and sediment yield and the published literatures have provided more details that are also quite relevant to our study. For example, plant growth can reduce raindrop energy and total runoff depth through canopy interception and stemflow(Wei et al., 2007; Ná var, 2011; Yang et al., 2013). Vegetation recovery can improve soil conditions, such as soil permeability and soil water storage capacity and can also reduce runoff production through root-network development and litter accumulation (Gyssels et al., 2005; Descheemaeker et al., 2006; Huang et al., 2013). Our study sufficiently demonstrated that vegetation restoration can effectively conserve both water and soil. For example, the runoff production was 13.55-fold lowerduringRSLD than during ESLD under forest conditions (e.g., Plot P). Our study (see Table 3) confirmed the earlier reports that the vegetation restoration became increasingly important for reducing runoff production once topographical features were stabilized and that rainfall duration and rainfall intensity were the dominant factors influencing runoff production under fully-restored vegetation condition(Wei et al., 2007, 2009, 2014).

The importance of the major factors affecting sediment yield also differed significantly between ESLD and RSLD. The descending order of the importance during ESLD was as follow: soil> rainfall> vegetation> topography> runoff(see Table 3). The descending order during RSLD was as follow: rainfall> soil> runoff> topography> vegetation. The published literatures demonstrated that removal of the vegetation and disturbance of the surface soil can degrade soil structures, thusencouraging sediment movement (Vacher et al., 2003; Porto et al., 2009). Some scholars have shown that the soil surface can be stabilized witha vegetation coverage> 60% and thus reduce soil erosion (Martin et al., 2010; Pohl et al., 2012). Our study further elaborated those points. For example, the sediment yield was 3.13-fold lowerduringRSLD than during ESLD under forest conditions (see Table 3).

As stated earlier, significant differences were observed under the different vegetation types in terms of runoff production and sediment yield betweenESLD and RSLD, being consistent with the reportedobservations by other scholars(Fu et al., 2000; Vacca et al., 2000; Meng et al., 2008; Xu et al., 2008; Xu et al., 2009; Mohammad and Adam, 2010; Wang et al., 2011; Ai et al., 2013; Wei et al., 2014). Although the runoff production and sediment yield differed significantly under different vegetation types, the variable coefficient for RSLD was lower than that for ESLDunder all vegetation types (see Fig. 4). It simply means that the effect of vegetation type on sediment yieldwas more important during ESLD than during RSLD. In addition, some scholars have attempted to search for the best plant species or the combinations for land restoration. For example, Wei et al. (2014) found that shrub species were better than grass species for retaining runoff and reducing sediment yield in a loess hilly area in China. This studyconcluded that RPa(H.rhamnoides+P.tabuliformis) and R (H.rhamnoides) are better choices for land restorationin this area, especially for relatively gentle slopes (i.e., less than 20 degrees).