Soil water response to precipitation in different microtopographies on the semi-arid Loess Plateau,China

Huan Ma·Qingke Zhu·Weijun Zhao

Abstract Soil water is an important factor restricting afforestation on the semi-arid Loess Plateau.The microtopography of the loess slope has changed the distribution pattern of soil water on the slope.To improve water utilization efficiency and optimize afforestation configuration patterns,the relationship between soil water and precipitation at micro-topographic scale must be studied.We used time series analysis to study the temporal variation of soil water and its response to precipitation in four kinds of micro-topographies and undisturbed slope on loess slopes.Micro-topographies significantly influenced soil water distribution and dynamics on the slopes.Soil water stored in the platform,sinkhole,and ephemeral gully influenced subsequent soil water for 4 weeks,whereas soil water stored in the scarp and undisturbed slope could influence soil water for 2 weeks. It took 12 weeks, 10 weeks,18 weeks,6 weeks,and 12 weeks for precipitation to reach the deeper soil layer in the platform,sinkhole,scarp,ephemeral gully, and undisturbed slope, respectively.These soil water characteristics in different micro-topographies are vital factors that should be taken into consideration when undertaking afforestation on the Loess Plateau.

Keywords Semi-arid Loess Plateau·Micro-topography·Afforestation·Time series analysis·Soil water

Introduction

Soil water is a key eco-hydrological variable in semi-arid areas(Ludwig et al.2005).Afforestation is an effective way to alleviate soil erosion on degraded land(Nunez-Mir et al.2015).On China's Loess Plateau,soil water is the primary consideration for site selection in plantations because of sparse natural precipitation(McVicar et al.2007).Seed germination and seedling growth require more soil water,especially during the early stages of afforestation.Site selection for plantations is a critical process for obtaining higher seedling survival rates during afforestation(Newton 2007).In 1985,research was undertaken on site type classification based on environmental factors on the Loess Plateau and soil water was found to be the main factor influencing the growth of Robinia pseudoacacia(Zhu 1985). Owing to a deficiency of soil water,afforestation projects implemented for controlling soil erosion have failed to achieve their anticipated goals in many cases(Jiao et al.2012).Even worse,poorly designed afforestation may lead to land degradation(Chen et al.2010)and soil desiccation(Guo and Shao 2003).Therefore,the study of soil water variation characteristics could better inform planning for site and species selection in afforestation.

In the past, the forestation model utilized uniform planting grids with the same row spacing along contours based on slope site conditions.This approach was adopted on the Loess Plateau for the sake of saving money and time cost. On the Loess Plateau, several micro-topographic variations have been shaped below slope scale because of water and gravity erosion(Zhao et al.2015).The microtopography alters the bearing surface of rain and the water migration path of the entire slope,which produces spatial heterogeneity in soil water and other habitat conditions.Soil water spatial heterogeneity causes the conventional afforestation model to lead to development of small,old trees(Zhu et al.2012).For more accurate knowledge of soil water features on the loess slope and for obtaining theoretical guidance on species selection in afforestation,studies regarding soil water should preferably be implemented at the micro-topographical scale.

The impact of topographic heterogeneity on ecological processes has long been recognized by ecologists(Dwyer and Merriam 1981).Topographic heterogeneity is known to affect several classes of response variables including the abiotic pattern(Gibson 1988),ecosystem process(Bubier et al.1993),and species distribution(Minchinton 2001).Researchers have studied micro-topographies mainly in wetlands(Courtwright and Findlay 2011;Gumbricht et al.2005).They found that micro-topographies significantly influence soil water(Chimner 1996),soil salinity(Domı´nguez-Cadena et al.2016),and soil nutrients(Moser et al.2009).On the Loess Plateau,great attention has been given to the influence of topographic heterogeneity on soil water distribution on slopes(Qiu et al.2001).Gao et al.(2016)studied the slope soil water in the Yuanzigou small watershed of the Loess Plateau and found that the presence of gullies clearly increased the spatial variability in terms of standard deviation,nugget,and sill.Yu et al.(2017)analyzed soil water in four different types of gullies and reported that different gullies exerted different influences on soil water behavior.Although the variation of soil water affected by topographic heterogeneity on the Loess Plateau has been extensively investigated, most research has focused on gullies,a topographic feature similar in extent to slopes.The smaller topographic features such as scarp,sinkhole and platform have not been adequately studied.

On the Loess Plateau,precipitation is the only source of soil water(Chen et al.2007).Time series analysis is an effective technique to characterize changes in soil water over time and to relate these changes to other observations(Nash et al.1991).Precipitation and soil water are strongly correlated over time but this relationship is affected by precipitation(Bergkamp 1998),topography(Jost et al.2005),soil characteristics(Wendroth et al.1999),and other factors.To improve the utilization rate of water resources in afforestation and to optimize tree species selection at the micro-topographic scale,we applied time series analysis techniques to:(1)quantify the variation of soil water content in different micro-topographies of pre-afforested loess slopes;(2)assess differential capacity for acquisition and short-term storage of soil water among micro-topographies; and (3) develop practicable guidelines for afforestation by applying results from the present study.

Materials and methods

Watershed description

This study was conducted in the 4.25 km2Hegou(36°54′09′′-36°54′23′′N,108°12′50′′-108°13′01′′E)watershed in Wuqi County,northern Shaanxi Province.The area is characterized by a temperate continental monsoon climate,which is warm and dry.The annual mean temperature is 7.8°C.The annual temperature varies greatly,with a maximum temperature of 37.1°C and a minimum temperature of -25.1°C. Annual mean precipitation is 467 mm(based on data spanning 1957-2012),with uneven distribution during different seasons.More than 85%of precipitation falls from May to October and the rainfall intensity during this period is very high(Fig.1).Minimum and maximum annual precipitation are 349 and 756 mm,respectively.During the study period of 2010-2012,mean annual precipitation was 557.2 mm. The underground water level is about 45-80 m deep.The soil type of this area is cultivated loessal soil and its thickness is up to 80 m.Loessal soil is a kind of young soil formed by direct cultivation of loess parent material.The loessal soil is loose and soft,thus the land surface can be easily eroded and form many types of micro-topographies owing to the intensity and duration of precipitation events.Since 1998,the Chinese government has implemented a policy of returning farmland to forests in this region to prevent soil erosion.The Hegou watershed has been under a grazing ban since then.At present,the main vegetation type is herbs with few trees and shrubs growing in gully bottoms.

Fig.1 Average annual precipitation of study area(1957-2012)

Classification of micro-topography

Micro-topographies on the Loess Plateau are generally shaped by water and gravity erosion.We focused on the micro-topographies located above the gully boundary line because most afforestation on the Loess Plateau occurred there.Based on field observations,slopes above the gully boundary on the Loess Plateau can usually be divided into disturbed slope and undisturbed slope(US)according to whether there is water erosion.On the disturbed slope there are four main kinds of micro-topographies that are partly formed,viz.scarp(SP),ephemeral gully(EG),sinkhole(SH),and platform(PL)(Fig.2)(Zhu et al.2012).SP is an area that has a substantially steeper cross-sectional profile than other parts of the slope;EG is the primary stage of gully erosion development,which is larger than a rill and smaller than a gully;SH is a crater on the ground shaped by running water;and PL is an area where the gradient is gentler than the other parts of the slopes.

Data collection

Micro-topographies were selected on the same slope and with the same slope-aspect(Table 1).A nearby site on an undisturbed slope where no micro-topography had formed was chosen as reference site.A 1.3 m long polyvinyl chloride(PVC)tube was pounded vertically into the soil at the center of each micro-topography and on the US for long-term observations.From each sampling tube,soil water content(SWC)was measured using a portable time domain reflectometry(TDR)system(TRIME-HD).The TDR system uses high frequency detection waves that can penetrate PVC tube to detect volumetric soil water content.SWCvalues were measured at 20 cm intervals along a 0-130 cm soil profile using a TDR probe.The probe took three measurements at each layer after rotating horizontally 120°and the means were used as the SWCof the layer.Soil water data were collected over 2 week intervals.A total of 42 observations were collected from August 2010 to June 2012.

Fig.2 Classification of upslope micro-topographies on the Loess Plateau

Precipitation data were collected at 1 week intervals using funnel-type rain gauges that were placed on top of the slope to eliminate wind effects.To have an equally spaced time series,precipitation and water content measurements were recorded at 2 week intervals.Precipitation values were summed over 2 weeks.The precipitation time series consisted of 42 observations.

Data analysis

Soil water storage

The TDR system recorded volumetric soil water content(SWC)while soil water storage(SWS)was also calculated for time series analysis.Soil water storage in each soil layer was calculated as:

where,Swsiis the soil water storage in layer i,Diis the thickness of soil layer i,and Swciis volumetric soil water content of layer i.The soil water storage over the total depth(TD)is the sum of Sws in each soil layer.

Time series analysis

We used an autocorrelation function to estimate the precipitation and soil water series and then calculate the crosscorrelation function to assess the response of soil water series to precipitation.

The autocorrelation function of time series z at lag h was calculated using formula 3 below(Box et al.2015):

where C （h） is the estimated auto-covariance that can obtained by the formula 4:

where¯z is the sample mean of the time series.

For two stationary time series x and y,if they collect data equally spaced in time,the cross-correlation function at lag time h can be estimate by formula 5 below(Box et al.2015):

Table 1 Main characteristics of sample plots

where,τxyh is the cross-covariance function of series x and y at lag h,τxx0 and τyy0 is the variance of series x and y,respectively.σxand σyis the standard deviation of x and y,respectively.

The cross-covariance function of series x and y at lag h can calculated using formula 6 below:

If the estimated value of the cross-correlation function ρxyh=0,the time series x and y were significantly irrelevant.If the estimated value of the cross-correlation function ρxyh0,the cross-correlation function value should compare with two standard errors that obtained from the Bartlet formula(Bartlett 1946).If the cross-correlation function ρxyh exceeds two standard errors,there is significant correlation between two time series(Box et al.2015).

Results

Dynamics of soil water and precipitation

The time series of precipitation for 3 layers of SWC,and TDof soil on the US are shown in Fig.3.The peak in soil water caused by precipitation first appeared in the surface soil(0-20 cm),and the peak in soil water in the 60 cm soil layer and 120 cm soil layer were 2 weeks and 4 weeks later,respectively(Fig.3).The peak time of the TDtime series in the 120 cm soil layer basically had a similar pattern to that of the 0-20 cm soil layer,which indicates that the variation in the surface soil water dominated the variation in TDafter precipitation.

Mean and coefficient of variation for the SWCof each soil layer and for the TDof the micro-topographies and US are listed in Table 2.Average soil water stored in the 120 cm soil layer in the micro-topographies and US differed significantly(F=8.044,df=4,P ＜0.001,oneway ANOVA).Total soil water at the 120 cm depth stored in SH was the highest(182.4 mm)whereas that stored in SP was the lowest (134.9 mm). The average TDwas SH ＞PL ＞EG ＞US ＞SP.These five terrain types have distinct varying patterns with increasing soil depth.SWCof PL and SP decreased with depth of soil,and then increased.In EG and US,SWCdecreased with soil depth.On SP,SWCinitially increased and then decreased.

Coefficients of variation(Cv)for all four micro-topographies and US increased with soil depth and then decreased.The maximal Cv of PL was at 40 cm while the other micro-topographies were at 60 cm. US had the highest Cv both in TDand average,which indicates that soil water in US varied dramatically compared with micro-topographies,and that US will not benefit for stable growth of seedlings.SP had the lowest Cv in TDand SH had the lowest average Cv.

Soil water response to precipitation

Autocorrelation of precipitation and soil water on the undisturbed loess slope

The estimated autocorrelation for precipitation,SWC,and TDon US performed differently(Fig.4).The autocorrelation for SWCand TDdecreased gradually until they became smaller than the standard error limits.The autocorrelations for precipitation decreased immediately after lag zero and became smaller than the standard error limits except for lag 1.

The time at which the standard error limit is reached is denoted as the sample ‘‘temporal correlation scale''(Box et al.2015).The range divides the sample into two groups.Observations within the range are correlated while observations beyond the range are essentially independent.The range of temporal correlation scales of soil water characterize the time interval that former soil water status can effect current soil water status.The range of temporal correlation scales for the 20,40,80,and 120 soil layers and for the TDwere at lag 0,1,2,2,and 1,respectively.This phenomenon indicates that deeper soil water had longer influence time to future soil water status.

Fig.3 Time series of precipitation,soil water content and total depth of water for undisturbed slope(US)

Autocorrelation of soil water in different microtopographies

To compare the autocorrelation of soil water in different micro-topographies,we summarized the range of temporal correlation scales for the four micro-topographies and US(Table 3).In general,the temporal correlation scale for all five topographical features and all depths(six soil layers)varied within 2 lags(4 weeks).For PL and US,the range of temporal correlation scales increased with the depth of soil.For SH,the range of temporal correlation scales decreased with the depth of soil.For SP and EG,the range oftemporal correlation scales initially increased and then decreased.The range of temporal correlation scales in TDindicates that soil water stored in US and SP can affect later soil water in 2 weeks,but the other micro-topographies can continuously impact later soil water for 4 weeks.

He made the youth follow him through dark secret passages, underground vaults22, and grey rocks till at last they came to an open field, which looked as if it belonged to a more beautiful world than ours

Table 2 Mean and coefficient of variation of soil water content and total depth of soil water storage for 4 kind of microtopographies and undisturbed slope

Cross-correlation of precipitation and soil water in microtopographies

The cross-correlations between precipitation and soil water on US are shown in Fig.5.The cross-correlations behaved smoothly with the lag time and were small at lag 0.The cross-correlations were positive at lag 0 and increased to their highest positive value for depths 0-120 cm and TD.Figure 5 also illustrates a definite periodicity in crosscorrelations.

The starting lag number represent the lag time when SWCresponded to precipitation and started to increase.The ending lag number represents the periodicity in the crosscorrelation,which means peaks and depressions in precipitation will be followed by peaks or depressions in SWCat intervals equal to the ending lag number.Table 4 summarizes the starting and ending lag numbers for the positive cross-correlations in the four micro-topographies and US.

Surface soil water(20 cm)in all micro-topographies and US started to respond to precipitation within 1 lag time(＜2 weeks). However, precipitation affected different depths of soil water among the four micro-topographies and US.Within lag 1,soil water at 80 cm depth responded to precipitation at PL,which indicates that rain falling on PL can infiltrate rapidly;the corresponding depth in SH was 60 cm whereas SP and EG were 40 cm.US has the shallowest soil depth response to precipitation during any given time frame.Precipitation reached to 100 cm depth in all micro-topographies on US,whereas the deeper soil layer(120 cm)reacted differently among the topographical features.The ending lags of cross-correlations for PL,SH,SP,EG,and US were 6,5,9,3,and 6,respectively.

The maximum significant cross-correlations between precipitation and soil water with the corresponding lag time numbers are given in Table 5.The maximum cross-correlation coefficient between precipitation and SWCindicates that the effect of precipitation on soil water was maximized at the corresponding lag time.Table 5 shows that the corresponding lag time of maximum cross-correlations of micro-topographies and US performed differently,with the lag time of maximum significant cross-correlations between precipitation and TDin PL,EG,SH,SP and US being 4,4,5,5,and 8,respectively.

Discussion

Soil water variations in micro-topographies

Fig.4 The correlograms for the precipitation and soil water content for the undisturbed slope;dashed line represent two standard error limits.(Each lag time equals 2 weeks)

In semi-arid areas,topographic heterogeneity can significantly influence soil water(Bergkamp 1998).The effect of soil water on vegetation growth is significant on the semiarid Loess Plateau(Hu et al.2009).It can help to understand the variations of soil water in micro-topographies when planning afforestation.In the present study,SH and PL had better soil water condition.For example,soil water storage in SH at 120 cm depth was 47.5 mm greater than that in SP.This indicates SH can be used as natural planting spots in afforestation.Our results are consistent with the previous findings(Yu et al.2017).However,the highest soil water of Yu et al.(2017)was found in the PL,whereas that of our study was in the SH.This difference resulted from the fact that Yu's study was conducted inmicro-topographies covered with trees but our study was conducted on pre-afforestation land vegeted with herbs.This indicates that vegetation could affect soil water in addition to the effect of micro-topographies.The interactive effect of micro-topography and tree species selection should be taken into account in afforestation planning.

Table 3 Significant autocorrelations with the corresponding lag time(range of temporal correlation scale)of 4 kind of micro-topographies and undisturbed slope.(Each lag time equals 2 weeks)

That SP had the lowest Cv in TDsuggests that soil water performed more stably on scarp than on other micro-topographies.However,previous study on the Loess Plateau showed that the Cv of soil water can decrease with declining SWC(Hu et al.2005).Due to the influence of the lowest recorded SWC,SP would not be a favorable microtopography for afforestation.

Precipitation variations in the study area

The autocorrelations of the time series indicated that current values were affected by previous values.We found that the observations of precipitation within 1 lag time(2 weeks)were correlated.This differs from the result found by Wang et al.(2007)on sloping land in the red soil hilly region in China where precipitation was independent at all lag times.This might have been caused by the specific precipitation feature of the area in the present study,where precipitation events occur intensively in dry and wet seasons and are distinct from each other.This feature suggests that future studies could focus on soil water contents of varied micro-topographies in both dry and wet season to more accurately define soil water characteristics.

Micro-topographies soil water short-term storage capacity

Autocorrelations of soil water meant that the current value of SWCwas affected by its previous value.A longer temporal correlation scale indicates better water retaining properties of the soil.The soil water retaining property is equivalent to the soil water supply capacity during the dry season in the semi-arid Loess Plateau.We found that PL had better soil water retaining property than other microtopographies and US.The temporal correlation scale of the majority of soil layers at 120 cm depth was at a lag time of 2(4 weeks).In contrast,the US and SP showed relatively poor soil water retaining properties.The temporal correlation scale at 20 cm depth for US and 120 cm depth for SP were less than lag 1,indicating that the soil water in those two topographies varied dramatically.

The antecedent soil water status is an important factor in hydrological modelling(Go´mez-Plaza et al.2000).Our results showed that an intensive observation interval of less than 2 weeks was necessary for obtaining the temporal variation of soil water on US and SP to build more accurate hydrological models.

Micro-topographies soil water acquisitive capacity from precipitation

Cross-correlation between precipitation and soil water demonstrates the water obtaining performance of soil.On the Loess Plateau,water loss exceeded precipitation in most months during the growing seasons because of canopy interception,soil evaporation,plant transpiration,and surface runoff(Jian et al.2015).Wang et al.(2013)reported that soil water was replenished mainly by 3-4 heavy precipitation events even during the rainy season.Our results showed that the precipitation could not recharge the deep soil layer(120 cm)in the semi-arid Loess Plateau over a short time(Table 4).The mean SWCin Table 2 also confirmed this phenomenon,which could lead to the variation of soil water in the topsoil dominating the variation in the total water depth.The reasons for this might be as follows.(1)Precipitation first supplied the surface soil owing to antecedent deficiency of water.(2)Hortonian overland flow frequently occurred in this area;therefore,plenty of water would be lost from runoffowingto high intensity precipitation(Chen et al.2016).(3)High temperatures lead to high evaporation because the wet season and summer overlap(Jian et al.2015).These factors ensured that the surface soil water was available for a relatively long period.However,our results showed that the four different types of micro-topographies could respond to precipitation quicker than did US.Especially in PL,precipitation could supply soil water at depth 80 cm within 1 lag time(2 weeks),which showed a remarkable capacity for obtaining water.Apart from SP,the other three micro-topographies in the present study had advantages in water storage and supply,which led to significant differences in SWC.

Fig.5 The cross-correlation between precipitation and soil water for the undisturbed slope.Dashed line represent two standard error.(Each lag time equals 2 weeks)

Implications for slope afforestation

Soil water characteristics of different micro-topographies should be considered when planning afforestation patternsand tree species selection on the Loess Plateau.Harrington(1991)found that soil water significantly increased biomass in semi-arid land.Hence,species that require more water required and are more productive in micro-topographies should be planted in better soil water conditions such as PL and SH.At present,the afforestation techniques commonly implemented on the Loess Plateau include runoffforestry,catchment irrigation,seed coating,and so on(Zhu et al.2012).However,these forestation models utilize uniform planting points with the same row spacing along contours based on site conditions of slopes.Our findings demonstrate that micro-topography should be taken into account when undertaking afforestation on the Loess Plateau area to improve seed germination and seedling survival rates.

Table 4 Positive lags for positive cross-correlation between the rainfall and water content for 4 kind of micro-topographies and undisturbed slope

Table 5 Maximum significant cross-correlation between the rainfall and water content for 4 kind of micro-topographies and undisturbed slope with corresponding lag time(each lag equals 2 week)

In the present study,soil water concentrations in PL and SH were the highest of the five topographical features and had ascendant capacity for obtaining water.Tree species that require more water should be selected for these two micro-topographies.Nevertheless,SH performed worse for water retaining property;therefore,Pinus tabuliformis and Populus hopeiensis should be selected for PL,whereas Ailanthus altissima and Hippophae rhamnoides should be planted at SH.For EG that had moderate SWSand inferior water retaining property,shrubs that can resist soil erosion such as Hippophae rhamnoides and Caragana korshinskii should be selected.For US that had the lowest SWCand soil water varied dramatically,herbs and small numbers of shrubs should be planted.For SP that had the lowest capacity for obtaining and retaining water,we recommend no planting in favor of natural restoration.On the Loess Plateau,utilization of land preparation to improve slope water use efficiency in afforestation has achieved effective results(Li et al.2011).However,artificial land preparation consumes a great amount of human power and material resources.Afforestation that utilizes the characteristics of micro-topography, which have similar effects to land preparation will save money and time,and will improve afforestation quality.

Conclusion

The differences in SWCbetween different micro-topographies and US were significant,with SWSranked as SH ＞PL ＞EG ＞US ＞SP. Precipitation redistribution on slopes led by micro-topographic features was a vital influencing factor on soil water variations among microtopographies and US.The temporal auto-correlation scale of the TDin PL,SH and EG was 4 weeks,and the temporal correlation scale of SP and US was 2 weeks.On the semiarid loess slopes,the effect of precipitation on the deeper soil layer(120 cm)was a relatively slow process:It took 12 weeks for PL,10 weeks for SH,18 weeks for SP,6 weeks for EG,and 12 weeks for US,respectively.Our results revealed that different responses to precipitation led to soil water variations in micro-topographies. These micro-topographic hydrological characteristics should gain increased attention in afforestation site and species selection.

AcknowledgementsThe authors would like to extend their sincere gratitude to Zongkai WU and Guangliang LIU for their help in providing local support during the study and Yingying ZHANG and Wenjuan MA for their help in reading and correcting proof.The author thanks Jutao ZHANG for the useful suggestions to improve the manuscript.The author also special thanks to editors and anonymous reviewers for their constructive comments to improve the manuscript.