Soil moisture and evapotranspiration of different land cover types in the Loess Plateau , China

We studied the impacts of re-vegetation on soil moisture dynamics and evapotranspiration (ET) of five land cover types in the Loess Plateau in northern China. Soil moisture and temperature variations under grass ( Andropogon), subshrub ( Artemisia scoparia ), shrub (Spiraea pubescens ), plantation forest ( Robinia pseudoacacia ), and crop ( Zea mays) vegetation were continuously monitored during the growing season of 2011. There were more than 10 soil moisture pulses during the period of data collection. Surface soil moisture of all of the land cover types showed an increasing trend in the rainy season. Soil moisture under the corn crop was consistently higher than the other surfaces. Grass and subshrubs showed an intermediate moisture level. Grass had slightly higher readings than those of subshrub most of the time. Shrubs and plantation forests were characterized by lower soil moisture readings, with the shrub levels consistently being slightly higher than those of the forests. Despite the greater post-rainfall loss of moisture under subshrub and grass vegetation than forests and shrubs, subshrub and grass sites exhibit a higher soil moisture content due to their greater soil retention capacity in the dry period. The daily ET trends of the forests and shrub sites were similar and were more stable than those of the other types. Soils under subshrubs acquired and retained soil moisture resources more efficiently than the other cover types, with a competitive advantage in the long term, representing an adaptive vegetation type in the study watershed. The interactions between vegetation and soil moisture dynamics contribute to structure and function of the ecosystems studied.


Introduction
In arid and semi-arid regions, water represents the main ecological constraint for plant survival, and hydrological processes determine the direction of evolution and ecological functioning of soil-vegetation systems (Li, 2011).Therefore, understanding the relationship and coupling mechanisms among soil, water and vegetation interactions can help to understand the land surface development processes and biogeochemical balances in the dry lands.
Soil moisture dynamics are the central component of the hydrological cycle (Legates et al., 2011) and are mainly determined by processes including infiltration, percolation, evaporation and root water uptake.Obtaining accurate estimates of root water uptake and vegetation water use represents the weakest link in producing soilvegetation-atmosphere transfer (SVAT) models (Schymanski al., 2008a).Optimality-based models have experienced rapid development in recent years and have shown strong predictive power, leading to the hypothesis that vegetation has developed optimal water use strategies through co-evolution with natural conditions, including reducing water losses passively and increasing water absorption capacities actively (Cowan and Farquhar, 1977;Schymanski, 2008).Schymanski et al. (2007Schymanski et al. ( , 2008a, b, c) , b, c) successfully reproduced the surface soil moisture dynamics using an optimality-based model and further tested it in catchments with natural vegetation in Europe.These model results show that the natural vegetation has adapted its water use strategies to local conditions.
The Loess Plateau of China is located in the upper and middle reaches of the Yellow River.It is a transitional zone between the southeastern humid monsoon climate and the northwestern continental dry climate.Almost all of the natural vegetation on the Loess Plateau has been replaced for cultivation as farmland in the last century.In the past three decades, to control soil erosion, a series of large reforestation campaigns was initiated, such as the Grain-for-Green project, which began in 1999.Indeed, approximately 24 % of the area of erosion has been controlled, and vegetation coverage increased from 6.5 % in the 1970s to 51 % in 2010 on the Loess Plateau (National Development and Reform Commission, 2010).Most of the farmlands cultivated on slopes were planted with forests and shrubs, and some farmlands were abandoned as grass and subshrub communities dominate (Su et al., 2011).
However, concerns have been raised regarding the ecohydrological effects of these artificial forests and shrubs that have induced unwanted environmental problems.Largescale vegetation restoration has also aggravated water scarcity, gradually leading to soil desiccation (Shangguan and Zheng, 2006;Chen et al., 2008a).Low forest productivity and growth efficiency resulted in "small aged trees" with heights of 3-5 m, a common scene in the region (Chen et al., 2008b).Vegetation-soil moisture feedback may lead to pattern formation (D'Odorico et al., 2007;Fu et al., 2011).Thus, understanding the interactions between these artificial vegetation types and soil moisture is urgently required as basis for adjusting land use structures and ensuring sustainable provisions of desired ecosystem services in this area.
The ecohydrological effects of vegetation restoration in the Loess Plateau area have been reported (Gong et al., 2006;Li and Shao, 2006;Sun et al., 2006;Chen et al., 2007aChen et al., , 2010;;Chen et al., 2008a, b).Many investigators have paid a great deal of attention to soil desiccation resulting from the excessive depletion of deep soil water by artificial vegetation and long-term insufficient rainwater supplies (Li, 1983;Yang, 1996;Yang et al., 1998;Li and Shao, 2001;Mu et al., 2003;Fan et al., 2004;Yang and Tian, 2004).Chen et al. (2007a) measured the soil moisture, runoff and soil erosion in plots of five vegetation types and found that soil water was lost during the rainy season and was not fully replenished from rainfall in a shrubland and a semi-natural grassland with moderatehigh moisture contents.The xylem sap flow in three species of trees was monitored in the field by Du et al. (2011), who found that the species vary in water use strategies, not only between exotic and native species, but also between the two native species.The number of studies on soil moisture spatial variability and temporal stability in the Loess Plateau continues to increase (Hu et al., 2009;Gao et al., 2011).These studies illustrated the water patterns for different vegetation species.However, the dynamics of soil moisture processes under different vegetation cover types are not clear, especially regarding the differences of soil moisture depletion and its vertical distribution.The main aim of this study is therefore to understand the soil moisture and water loss patterns under different land covers by continuously monitoring soil moisture fluxes of grass (Andropogon), subshrubs (Artemisia scoparia), shrubs (Spiraea pubescens), plantation forest (Robinia pseudoacacia), and crops (Zea mays) during and after rainfall events.

Study area
We chose four typical land cover types with a similar slope position, aspect, and slope degree, including plantation forest, shrub, subshrub, grass species and a crop site on the check dam lands in the bottom of the valley, in the Yangjuangou Catchment (36 • 42 N, 109 • 31 E), located in the central part of the Loess Plateau (Fig. 1).Driven by the implementation of the Grain-for-Green project since 1999, most of the cultivated lands on steep slopes in this catchment were gradually abandoned for natural and artificial re-vegetation.Therefore, a mosaic of patchy land cover is the typical landscape pattern in the area, and the current main vegetation types were formed during different restoration stages associated with varying soil conditions.This region represents a typical hilly geomorphology with well-developed gullies, with elevations ranging from 1050 m to 1298 m and slope gradients between 10 • and 30 • (Liu et al., 2011).The area has a semi-arid continental climate, in which the mean annual precipitation and air temperature during the past 20 yr (1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007) were 498 mm and 10.6 • C, respectively.The rainfall is mainly concentrated between June and September, with large inter-annual variations, and spring and early summer are usually characterized as a dry season.The growing season for the common deciduous tree species ranges from April to October in this region.The soil in the study area is mainly derived from loess, with a depth of 50-200 m depending on topography.The loess in this area usually exhibits a texture consisting of more than 50 % silt (0.002-0.05 mm) and less than 20 % clay (< 0.002 mm), with a porosity of approximately 50 %.The gravimetric field capacity and wilting percentage of the soil water in the region are 20-24 % and 3-6 %, respectively (Du et al., 2011).
The forest site was dominated by Acacia (Robinia pseudoacacia) with a tree height of approximately 5.5 m; the distance between rows was 2.5 or 3 m.The shrub site was densely covered with Spiraea pubescens with a height of approximately 1.5 m and a sparse layer of planted acacia.The subshrub site was covered with Artemisia scoparia with a height of approximately 0.45 m imbedded with tussock and bare areas.The grass site was covered with Andropogon beard grass with a height of approximately 0.40 m with imbedded bare areas.The crop site was situated on dam land at the bottom of the valley and was cultivated with corn (Zea mays) with a height of approximately 2.20 m (Table 1).The grasses and subshrubs grew under conditions of natural succession.The slope gradients of the sites were approximately 22 • , with only slight differences, except for the crop site.

Field installation
A total of 12 soil moisture and temperature smart sensors in every site were installed at 6 depths below the ground at 10, 20, 40, 60, 80 and 100 cm.To install the probes, a pit was dug in the soil with a sufficient width to allow their insertion.The probes were inserted into the soil through the unaltered side of the pit and were positioned horizontally in the direction of the maximum slope of the terrain.Once the probes had been inserted, the pit was carefully refilled, avoiding perturbations as much as possible, and the surface was contoured in a manner similar to the surrounding slope.The site was set up at the end of April 2011, and measurements were not begun until 2 months later, when the soil settled.

Measurement sensor
H21 Soil moisture & Temp Logger Systems with S-SMC-M005 soil moisture probes and S-TMB-M006 soil temperature probes (Decagon Devices Inc., Pullman, WA) were installed to measure the soil profile moisture and temperature.The S-SMC-M005 soil moisture sensor is capable of measuring volumetric saturations between 0 % and 100 % with an accuracy of ±1.0 %, and data were collected by HOBO weather station logger every 10 min.Rainfall was measured using a tipping bucket rain gauge, which was connected to a data logger with a precision of 0.2 mm.Concurrently, other meteorological parameters, such as the air temperature, relative humidity, wind velocity and potential evapotranspiration (ET 0 ), were recorded at a height of 2 m above the ground every 30 min.

Data analysis
We assumed that the volumetric soil water content was accurately measured with probes matching the different depths in the soil profiles.Based on the soil water balance principle, the cumulative loss water can therefore be described as follows: where S i is the initial volumetric soil water content (cm 3 cm −3 ); S e is the volumetric soil water content (cm 3 cm −3 ) at the end stage; Z f is the measure depth (cm); and L is the cumulative loss water (mm).Data on soil moistures of these five types were analyzed using SPSS for Windows 16.0.A one-way ANOVA was  performed, after verifying the assumptions of normality and homogeneity of variances, to test the effects of land cover types on soil moisture.Multiple comparisons were made using the least significant difference (LSD) method.

Soil moisture pulse
Over the entire duration of the study, from June to September, there was no obvious variability in the sensor readings obtained at the 80 and 100 cm depths, and the average of the other four sensor readings under each land cover type was therefore used.Based on the responses of the moisture probes to rainfall events, there were more than 10 moisture pulses during the period of data collection (Fig. 2).In all of the rainfall episodes, a similar parallel, but different trend in moisture retention and reduction is noticeable between the five investigated land cover types.Typically, the variability in soil moisture readings was correlated with the amount of precipitation received.The highest soil moisture content peak was experienced on 20 August due to three major rainfall events in the three preceding days.Throughout the observation period, the surface soil moisture showed an increasing trend.

Comparison of soil moisture and temperature in different land cover types
The corn cover was associated with consistently higher soil moisture readings than the other surfaces (p < 0.01).The grass (Andropogon) and subshrub (Artemisia scoparia) sites showed an intermediate level, with the values for grass being slightly higher than those for subshrub most of the time until 16 September (p = 0.26), significantly different from others (p < 0.01).The shrub (Spiraea pubescens) and plantation forest (Robinia pseudoacacia) sites presented lower soil moisture readings, with the shrub values being slightly higher than those of the forest site consistently (p = 0.01), and significantly different from others (p < 0.01) (Table 2).With exception of the corn land cover type, the average soil temperature was clearly affected by the growth of foliage.The average soil temperature exhibits approximately the same regime as the average moisture content among the other four cover types.The temporal trend of the average soil temperature contrasted with the moisture trend; they are significantly negative correlated for all these five types (p < 0.01), and show a downward trend throughout the observation period (Fig. 3).

Pattern of soil moisture decreases
Three typical decreasing moisture periods were chosen to compare the differences in the water loss rate between the five land cover types and their vertical distribution.
In the period of relatively lower soil moisture contents, from 6 July to 12 July, the initial average soil moisture content was 0.15 cm 3 cm −3 .The subshrub site lost the greatest amount of water of up to 15 mm over these 7 days, which was mainly contributed by the 0-10 cm and 10-20 cm layers.Additionally, the daily water loss was higher for subshrub than for the other cover types, ranging from 0.7 to 2.6 mm.The lowest total water loss value over this 7-day period of only 6 mm was observed for grass cover, which showed a daily water loss trend that was similar to but lower than that of subshrub, ranging from 0.15 to 1.09 mm.The daily water loss for corn was the most variable, ranging from 0.38 to 2.43 mm, and the cumulative water loss for this cover type was 7.8 mm.Plantation forest and shrub showed an intermediate level, with the value for the forest site being slightly higher than that of shrub.These last two sites showed cumulative water losses of 9.5 and 7.2 mm, respectively, and their daily water loss trends were similar and more stable compared to the other cover types.The water losses of these four cover types were mainly contributed by the 0-10 cm layer in this period (Figs. 4 and 5).
In the periods of relatively higher soil moisture content, from 31 July to 15 August (Figs.6 and 7) and 25 August to 31 August (Figs.8 and 9), the initial average soil moisture content for all was 0.22 cm 3 cm −3 .The data for the forest site were missing for 9 August, and those of the shrub site were missing from 8 to 15 August.During these two periods, the average daily water losses for subshrub and grass were 2.3 and 2.2 mm, respectively.Corn showed the lowest average daily water loss of 1 mm.The forest and shrub sites presented an intermediate level, exhibiting average daily water losses of 1.7 and 1.8 mm, respectively.All three layers contributed to the water loss among the five cover types.

Soil moisture dynamics
Soil moisture reflects the balance of precipitation, runoff, and ET and exhibited various types of pulse events.The precipitation events occurring on the Loess Plateau can be simply divided into two categories: events with small amounts of rainfall occurring at a high frequency and events with large amounts of rainfall with a low frequency.The amount of water received in the form of small events shows little variation, whereas the amount of water in large events varies markedly among years, leading to the large inter-annual variations observed in amount of precipitation in this area (Liu et al., 2011).Small rainfall events affect only the uppermost 1 cm of the soil, and the soil moisture was therefore mainly contributed by the large events.Infiltration, surface run-off and evapotranspiration have been identified as the key factors determining soil moisture content at fine scales.Evapotranspiration (ET) is a large term in the water balance equation in arid and semi-arid areas, including evaporation from soil and transpiration from vegetation, which are processes that  The significance is within the brackets.a represents significant correlation at level of 0.01.b represents significant correlation at level of 0.05.
usually cannot be separated (Wang et al., 2012).A large fraction of the soil moisture in the surface layer is lost through direct evaporation due to high temperatures and low root densities, while the rates of plant water uptake increase, and evaporation and vapor diffusion rates decrease in deeper soil (Schwinning and Sala, 2004).Thus, the larger the rainfall event is, the deeper the associated pulse depth and the larger the fraction of precipitation leaving the soil via transpiration and contributing to the primary productivity of higher plants.
The loss of soil moisture through the transpiration of higher plants, especially the woody plants with deeper root, is more stable than the evaporation from soil due to their constant water uptake capacity.Our observation period fell within the rainy season, when there is higher soil moisture.The subshrub and grass cover types cannot protect the soil surface from solar radiation, leading to greater daily water losses via direct evaporation.The interval of the rainfall events was not sufficiently long for the moisture loss regime to shift from the wet period to the dry period.Higher moisture retention capacity after rainfall events also implies higher moisture loss through direct evapotranspiration before a wet/dry threshold is reached, consistent with the results conducted in the Eastern Cape province, South Africa (Odindi and Kakembo, 2011).The larger amount of water that infiltrated under the subshrubs led to a higher moisture loss rate, whereas grass cannot maintain this high rate for as long a period due to the smaller amount of water acquired.The initial soil moisture content of the subshrub and grass sites was higher than that of forest and shrub, and the moisture loss rates for subshrub and grass were also higher than those of the forest and shrub sites in this short wet period, leading to the gap between them becoming smaller.The corn site, situated on check dam land in the bottom of the valley, received more water flow from hillslope, and the valley areas had lower temperature, resulting in their higher and more stable soil moisture content.

Soil temperature and water loss
Temperature is one of the main factors causing the loss of soil moisture under conditions of high moisture content, which is also influenced by leaf area index (LAI) and factors.In the wet period after a rainfall event, soil moisture loss is controlled by atmospheric demand or energy (S.Wang et al., 2011).The corn site, which was located on dam land, differed from the other cover types because of its terrain and soil properties.Among the other four types, subshrub and grass showed the highest soil temperatures, corresponding to greater moisture losses.The shrub site presented the strongest effect due to high coverage, resulting in lower temperatures and less moisture being lost during the wet period.The temperature at the plantation forest site was slightly higher than that of the shrub site, as observed for the moisture losses.These water losses in the surface layer may be mainly through the direct evaporation.According to Odindi and Kakembo (2011), moisture losses include both wet and dry regimes presenting different loss rates.The wet regime has a higher rate than that of the dry regime.The frequent rainfall events during our observation period prevented a full moisture decrease during the wet period.De Lannoy et al. ( 2006) also observed that a reduction in moisture content leads to a decrease in the rate of evapotranspiration, resulting in wetter patches experiencing more rapid soil moisture loss than other surfaces.Although the plantation forest and shrub vegetation consumed more water in the growing season, as the growing season and the rainy season coincide, the consumed water was supplemented during this period.The soil moisture gap occurred in the dry season when the soil moisture loss was mainly controlled by the capacity of plants to absorb water from the soil (S.Wang et al., 2011).Forest and shrub vegetation can maintain high and stable transpiration during the dry season, whereas subshrubs and grass are smaller, which with other factors contributed to the differences in the soil moisture content between them.

Implications for re-vegetation
The most important lesson regarding water and soil conservation in the Loess Plateau area of China is that "precipitation should be impeded to allow it to infiltrate locally" (Chen et al., 2007b) to reduce surface runoff and enhance infiltration.Following implementation of the Grain-for-Green project for more than 10 yr, most of the cultivated lands on the slope were planted with artificial forests and shrubs, though grasses, in some cases succeeding to subshrubs, were also grown on some of the abandoned cultivated lands.The effects of this re-vegetation regarding impeding surface flows are obvious, with runoff being reduced and the goal of soil and water conservation achieved.Unfortunately, because the amount of rainfall interception associated with forest and shrub vegetation is greater than crops, the high ET results in higher soil bulk density and the soil becomes firmer with lower infiltration capacity.Thus, the amount and depth of infiltration associated with plantation forest are lower than for grass soils.The higher evapotranspiration capacity associated with plantation forest and shrub vegetation results in a lower soil moisture content and even a drier layer (Chen et al., 2010).Sparely distributed subshrubs acquire and retain soil moisture resources more efficiently than plantations and dense shrub vegetation.This is due to the efficient impediment effects and medial root water uptake capacity of subshrub vegetation, which also gathered and maintained a high clay content.Hillslope plot experiments suggest that grass cover yields more runoff (Liu et al., 2011), leading to more sand and lower clay content.However, the grass site could retain more moisture due to its shallow root distribution and lower water uptake capacity.The amount of water loss of grass is determined by the initial moisture content, so grass cover will develop into subshrubs, as the subshrubs have a competitive advantage on the long term.More than 100 000 check dams have been built over the last 50 yr in the Loess Plateau, which store 21 billion m 3 sediments, and 3200 km 2 of dam croplands have been created, and the carbon storage in check dams of the Loess Plateau can amount to 0.952 Gt (Y.F. Wang et al., 2011).However, dam lands exhibit a higher moisture content but a low permeability and a potential for salinization as well as higher bulk density and a clay layer in deep soils.

Conclusions
This study quantified soil moisture retention and flux variations under plantation forest, shrub, subshrub, grass and corn cover in the re-vegetated catchment area of the Loess Plateau in the rainy season of 2011.Despite the greater postrainfall loss of moisture under subshrub and grass vegetation than forest and shrubs, subshrub and grass sites exhibit a higher soil moisture content due to their greater soil retention capacity in dry periods.The soil temperature exhibits the same regime as the moisture loss following rainfall events, with the exception of the corn site.Changes to land cover may alter the soil moisture budget, and soil moisture is one of the most important abiotic factors determining vegetation growth, variability and regeneration.The interaction between vegetation and soil moisture dynamics has important implications to structure and functioning of these ecosystems.Subshrubs have been the natural succession vegetation type following the abandonment of the croplands in the study area for more than 20 yr.The runoff and run-on patches associated with this vegetation type acquire and retain soil moisture resources more efficiently than other cover types.Thus, subshrubs represent an adaptive vegetation type in this area.

Fig. 1 .
Fig. 1.Location of the study area and the typical land cover types: grass (A), subshrub (B), shrub (C), plantation forest (D) and crop (E) site.

Fig. 2 .
Fig. 2. Dynamics of the mean soil moisture and precipitation profiles and the three selected typical decreasing moisture periods.

Fig. 3 .
Fig. 3. Dynamics of the mean soil and air temperature during the study period.

Fig. 4 .
Fig. 4. Soil moisture decrease process under the five land cover types from 6 July to 12 July.

Fig. 5 .
Fig. 5. Daily soil water loss (a) and cumulative water loss (b) under the five land cover types from 6 July to 12 July.

Fig. 6 .
Fig. 6.Soil moisture decrease process under the five land cover types from 31 July to 15 August.

Fig. 7 .
Fig. 7. Daily soil water loss (a) from 31 July to 15 August and cumulative water loss (b) under the five land cover types from 31 July to 7 August.

Fig. 8 .
Fig. 8. Soil moisture decrease process under the five land cover types from 25 August to 31 August.

Fig. 9 .
Fig. 9. Daily soil water loss (a) and cumulative water loss (b) under the five land cover types from 25 August to 31 August.

Table 1 .
Main characteristics for different land cover types.
a Mean value.b Standard error.

Table 2 .
The comparisons of average soil moisture between the five land cover types.