Effects of changes in soil hydraulic properties after crop abandonment on co-occurring 1 perennial species in a semi-arid grassland in Mongolia 2 3

18 The objective of this study was to investigate successional changes in water flow as a 19 result of changes in soil hydraulic properties after crop abandonment under drought and non20 drought conditions, and under water uptake by co-occurring perennial plant species to clarify the 21 observation that typical perennial grass species are seldom observed in abandoned fields. Soil 22 hydraulic properties were measured in croplands which had been abandoned for different periods (2, 23 9, and 18 years from abandonment) and in a grazed grassland site. Hydrological processes in the 24 soil profiles were simulated with soil hydraulic properties under drought and non-drought summer 25 conditions with water uptake from perennial grass species Suction in the surface soils increased 26 with the period of abandonment, with this trend being particularly obvious in a drought year. 27 Available water appears to be restricted in the later successional stage of abandoned fields and in 28 grazed grassland for plants that have drought tolerance. Dry soil and climate conditions are 29 important factors determining the intrusion of the typical perennial grass, S. krylovii, into degraded 30 abandoned fields. This abiotic interaction between soil hydraulic properties and climate conditions 31 may play an important role for plant succession in abandoned cropland. 32 33


Introduction
Crop abandonment is a factor responsible for soil degradation in semi-arid regions.
Soil degradation within a few decades following abandonment of cropping is a serious problem in Mongolia and Inner Mongolia [1,2].The dominant land use in these areas is grazing.Palatable grasses for livestock are originally dominant.Grasslands are dominated by the perennial grass, Stipa krylovii, which has lower vulnerability for drought than degraded grassland where annual species dominate [3].However, the dominant species of typical perennial grasslands are seldom found in abandoned fields in Mongolia, even decades after abandonment [1] and in Inner Mongolia [4].Instead of typical perennial species, annual species and the perennial species (Leymus chinensis), which has a high compensation rate, tend to dominate such grasslands [1,4].
Active ecological restoration has been tried in Mongolia to accelerate the recovery process.For instance, sowing of typical perennial grasses [5] and making gaps by removal of annual species [6] in abandoned fields have been tried.However, these activities have not brought the expected results due to unexpected drought and degraded soil conditions [5,6].These limitations of restoration activities may result from having inappropriate particular goals or from targeting apparent symptoms without considering underlying causes [7].Therefore rigorous assessment of the state of abandoned cropland in the Mongolian rangeland and the underlying factors leading to that state, such as changes in soil hydraulic properties and plant species composition under annual variability of precipitation, should be undertaken as a first step in restoration activity.
The effects of crop abandonment on soil restoration may depend on soil properties and climatic conditions of an area [8].In particular, soil hydraulic properties affect the vegetation recovery process [9].Soil water suction interacts with water uptake by plant roots.Therefore, measuring the changes in average suction after crop abandonment is critical in understanding soil water conditions for typical perennial plant growth.Previous studies have found changes in particle size distribution [1,10,11], hydraulic properties of the near-saturated range of surface and subsurface soils [12,13] and unsaturated hydraulic conductivity (i.e., high suction range) of surface soils [1] due to crop abandonment in the semi-arid region.However, there are insufficient studies on successional changes in infiltration into and capillary rise from the subsoil that occur when considering the changes in unsaturated hydraulic properties after crop abandonment.In addition to changes in hydraulic properties of the soil profile, there is substantial annual variability in climatic conditions in semi-arid regions.Therefore it is important to investigate soil hydraulic properties under various climatic conditions when evaluating the soil water condition in places where typical perennial species grow to answer the question as to why typical perennial species such as S. krylovii have seldom been observed in abandoned cropland.
The objective of this study is to present the successional changes in water flow due to changes in the soil hydraulic properties after crop abandonment under drought and non-drought conditions, and under water uptake by co-occurring dominant plant species to clarify the reason why typical perennial grass species are seldom observed in abandoned fields.

Study site and soil sampling
The study area was in Hustai National Park (HNP, 47°50′N, 106°00′E) in Mongolia.
The average elevation of HNP is 1240 m asl.The climate of the region is semi-arid and cold, with short summers.Most of the annual precipitation falls in summer, from May to July, which is critical for the growth of plants.According to meteorological data from the HNP climatic station, annual precipitation was approximately 232 mm (C.V. 31%) and the mean annual temperature was approximately 0°C from 1999 to 2005 at HNP.The zonal soils were classified as Kastanozems by WRB [14] based on soil profile morphology and physicochemical properties.
All abandoned croplands had been tilled from 1977 in this area by administrative plan.Satellite images (Landsat MSS 1994, 1996, 1999, 2002, 2005, 2007) were used to confirm the period of abandonment.We then consulted municipal archives and held personal interviews with landowners to determine the period of abandonment.We selected croplands that had been abandoned in 1990 (site CA18), 1999 (CA9) and 2006 (CA2) and semi-natural grassland (SNG) in the flat buffer zone in Hustai National Park as a reference site.The CA2, CA9 and CA18 sites had been abandoned from cropping two, nine and eighteen years, respectively, prior to our survey in summer 2008.Wheat had been cultivated prior to the cropland being abandoned.The cropping system and agricultural materials (e.g.fertilizer and agricultural machines) used at each site basically followed the government guidelines until 1990.Although the abandoned cropland sites had had different durations of cultivation, previous studies reported that half of the changes in soil physicochemical properties occur during the first 8 years from primary tillage, and subsequent changes are slow [2,15].All of the study sites of abandoned fields had been cultivated for more than 12 years.Hence, the study sites should allow proper comparison of the effects of crop abandonment on soil and plant properties between these sites.Grazing was introduced on all abandoned cropland sites after crop abandonment.Livestock numbers were strictly controlled by HNP in the buffer zone including CA2, CA9, CA18 and SNG.
Soil surveys and soil sampling were performed from July to August 2008.Soil cores (100 ml) were sampled from depths of 0-5, 10-15 and 30-35 cm in the three different fields to investigate the variation in each field.The selection of soil layers for sampling was based on the soil profile observations.The sampled soil cores did not contain the boundaries between soil horizons.

Calculations of plant growth and hydrological fluxes
Initial simulations of soil water flow were performed for the bare surface condition to focus on the effect of changes in soil hydraulic properties with different periods of crop abandonment and land-use history on average suction and volumetric water content.We then simulated the cumulative evaporation of the dominant two species (Stipa krylovii and Leymus chinensis) and water movement under those plants, to compare the effects of drought and differences in soil hydraulic properties on the growth of these two species.Since S. krylovii and L. chinensis were not observed at the CA2 site, we did not use the soil hydraulic properties of this site.
The various terms of water balance in one dimension were simulated using WASH_1D (http://www.alrc.tottori-u.ac.jp/fujimaki/download/WASH_1D/WASH_1D.zip), which solves the one-dimensional Richards' equation and convection-diffusion equation for heat movement with the finite difference method.The WASH_1D program considers the dependence of albedo on water content and vapor movement due to the thermal gradient and hysteresis.The potential evaporation rate was calculated by the Penman equation.A plant growth model is also involved in this program.
We simulated the soil water movement in the 0-5 cm, 5-15 cm and 15-50cm soil layers with core soil samples from the 0-5 cm, 10-15 cm and 30-35cm depths, respectively.Tables 1 and 2 show all parameter values for soil and plants, respectively.

Meteorological data
Meteorological data for precipitation, maximum and minimum temperatures, wind velocity, relative humidity, sunshine duration and direct radiation for every 3 h were used in the simulation.These data were obtained from the Institute of Meteorology and Hydrology (IMH) and World Radiation Data Centre in Tov Prefecture where HNP is located.Since we focused on soil hydraulic properties as the critical condition determining water availability for plants, the simulated period was restricted to the plant growth period (May-July).To reflect a typical climate difference, we selected data from non-drought (1999) and drought (2004) summers.
Amounts of precipitation during the relevant periods were 158 and 81 mm, respectively.We assumed that the initial suction was uniform at 300 cm.

Soil hydraulic properties
Saturated hydraulic conductivity was measured by the falling head method for every soil core sample.Unsaturated hydraulic conductivity and retention data were measured by the evaporation method using two tensiometers [16].Those at the surface layer were determined by an evaporation method, which gives hydraulic conductivity at a high suction (>600 cm) range.In this evaporation method, two tensiometers were used until suction at the upper tensiometer reached more than 500 cm.After removing the water in the tensiometers, the soil column was again placed under the evaporative condition until the evaporation rate became lesser than 10% of that at the initial stage.After termination, the soil columns were dismantled to obtain the water content profiles.Parameters in hydraulic conductivity functions in the high suction range were inversely determined using tensiometer readings, cumulative evaporation derived from the weight change of soil samples through the experiments and the water content profile in the objective function.The values for unsaturated hydraulic conductivity and retention curves were determined in this way and validated by comparison with unsaturated hydraulic conductivity and water retention data measured using the multi-step outflow experiment of Fujimaki and Inoue (2003) [17].We further validated the method that we used with the known hydraulic properties of the Kanto loam.Retention data in the high suction range (14077 cm, 402941 cm) were obtained using the vapor pressure method for surface core samples because soil water suction covers a wide range in the semi-arid Mongolian grassland.
The water-retention data were fitted with a modified version of Shiozawa's equation: where θ is the volumetric water content, θsat is the saturated θ value, h is the suction (cm), and α, ζ, m, n, and h0 are fitting parameters.The unsaturated hydraulic conductivity (cm h -1 ) data were fitted with Campbell's equation: where ak, bk, ck are fitting parameters.The relationship between water content and thermal conductivity (Kh) is given by: , where ah, bh, ch and dh are fitting parameters.We estimated these fitting parameters using the method proposed by Campbell (1985) [18] for the surface soil (0-5cm).The estimated thermal conductivity function was verified by comparison with that measured using a thermal conductivity probe (Decagon Devices Inc., Pullman, WA, USA).The same parameter values were used for subsurface soils.The dependence of albedo on the water content (R) is given by: where l is the average volumetric water content in the top l cm (cm 3 cm −3 ) and max, min, al and al are fitting parameters.We took measurements using a newly designed device and pyranometer (Ll200X Campbell Scientific, Inc) that requires only a small amount of soil sample and allows quick measurement across a wide range of water contents to obtain the parameter values of each soil sample.The accuracy of the values measured using this device was validated using the known data of Fujimaki et al. (2003) [19].

Calibration for simulation of soil hydraulic properties
To verify the accuracy of the predicted water movement, two soil moisture sensors (5TE, Decagon Devices, Inc.) were inserted to monitor the water content, electrical conductivity and temperature at depths of 5 and 15 cm at 3-h intervals, from July 24 to August 8, 2008.
Precipitation (ECRN-100, Decagon Devices, Inc) and shortwave radiation (LI200X, Campbell Scientific, Inc) at 3-h intervals were monitored in the same period.Other meteorological data (maximum and minimum temperature, wind velocity and relative humidity) were obtained from IMH.We took core soil samples from depths of 2.5-7.5 and 12.5-17.5cm from the point where the soil moisture sensors were inserted.Saturated and unsaturated hydraulic properties were measured by the falling head and evaporation methods.Retention data in the high suction range were obtained with the vapor pressure method for surface core samples.We estimated fitting parameters of thermal conductivity and dependence of albedo on the water content using the same method as for the other sampled surface soils.We compared the simulated data of the volumetric water content at depths of 5 and 15 cm using the sampled soil parameters and monitored the values of volumetric water content.The RMSE between simulated and monitored soil water content at depths of 5 and 15 cm were 0.0053 and 0.0181, respectively.

Plant properties
The actual transpiration rate, T, is computed by The sink term, S (s -1 ), is defined as the volume of water removed from a unit volume of soil per unit time due to water uptake by a plant, and z is soil depth (cm).Feddes et al. (1978) [20] specified S as where Tp,  and w are the potential transpiration rate (cm s -1 ), normalized root density distribution (cm -1 ) and reduction coefficient, respectively.Tp was given by multiplying the reference evapo-transpiration rate using the Penman equation [21] and crop coefficient (Kc).Kc is expressed as a function of cumulative transpiration: where akc, bkc and ckc are fitting parameters.The values of akc, bkc and ckc were assumed to be 0.8, -0.4 and 0.3, respectively, based on a previous study [22]. was given by: where b is a fitting parameter, and drt and z are the rooting depth (cm).The value of b was assumed to be unity, based on data from an experiment using these two plants, which means that normalized root density distributions are distributed linearly.The drt is also expressed as a function of cumulative transpiration: where adrt, bdrt and cdrt are fitting parameters.The values of adrt, bdrt and cdrt were assumed to be 25, -0.04 and 10, respectively, which means that the rooting depth becomes dipper-shaped depending on cumulative transpiration.These parameter values were adapted by assuming S. krylovii and L. chinensis have similar rooting depths in the early stage of making a habitat.The w was described using the S-shaped function: where h50 is the soil water pressure head at which uptake is reduced by half and p is an adjustable parameter.In this work, we set h50 = -1302, p = 1.355 and h50 = -1254 and p =4.862 for Stipa krylovii and Leymus chinensis, respectively, given by averaging determined parameters in our previous research [23] (Fig. 1).The parameters were obtained by the pot experiments of fullgrown S. krylovii and L. chinensis (Table 2).Present study, therefore, did not consider about germination process from seed in simulation.Leaf area index (I) was expressed as a function of cumulative transpiration.
where aLAI and bLAI are fitting parameters.Assumed values of aLAI and bLAI were 1 and 0.2, respectively.Parameters related to LAI were assumed based on the results of a previous study in S. krylovii and L. chinensis dominated grassland [24].Initial cumulative transpiration was 0.1 cm based on field survey results in Mongolia [25].Since shortwave radiation flux which arrives at the soil surface is decreased due to vegetation cover, the shortwave flux is expressed as a function of leaf area index.Evaporation rate (E) is calculated by a bulk transfer equation as follows: where vs is the saturated vapor concentration at the soil surface (g cm -3 ), va is the saturated vapor concentration at reference height (g cm -3 ), hrs is the relative humidity at the soil surface (g cm -3 ), hra is the relative humidity at reference height, ra is the aerodynamic resistance (s cm -1 ) and rsc, is the resistance due to salt crusting (s cm -1 ).An increase in aerodynamic resistance due to plant cover was also expressed as a function of leaf area index.Vegetation dependent parameters were filled with theoretical values except for h50 and p of each species, and each parameter was assumed to be the same value for every numerical experiment.water flow in the soil.This response function was the average of the responses of three plants of each plant species.

Particle size distribution
The soil core samples were air-dried and sieved to obtain all particles <2.0 mm after the hydraulic measurements.The particle size distribution for each site was determined using the pipette method.

Particle size distribution and hydraulic properties of surface soil samples
The average particle size distribution of soils at the different sampled sites is shown in Table 3.The coarse sand content was highest at the CA2 site, and its fine sand content was higher than at the cropland sites abandoned for longer periods.The fine sand content at the SNG site was the highest of all sites.
Figures 2 and 3 show the unsaturated hydraulic conductivity and water retention curves for the surface soils at the four sites.The unsaturated hydraulic conductivity of soils from abandoned fields, especially at the CA2 site, was higher than that for the soil at the SNG site at a given volumetric water content.The suction of the soil from abandoned fields at the CA2 site was lower than that for the soil at the SNG site at a given volumetric water content (ex. volumetric water content is 0.2 cm 3 cm -3 ).Suction values for soils at the CA9 and CA18 sites fell between values from the CA2 and SNG sites.

Volumetric water content and soil water suction in drought and non-drought summers
Figure 4 shows the average volumetric water content of surface soils from the different sampled sites in the summer of drought and non-drought years.There are no obvious differences among sites.Figure 5 shows changes in soil suction in the same years.There were large differences in terms of suction in the surface soil.Soil suction in abandoned fields was lower than that at the SNG site, and at the CA2 site, was the lowest of all sites in the drought year.Although there was not a large difference in suction between the CA9-and CA18 sites in the non-drought year (Fig. 4a), the suction at the CA18 site was higher than that at the CA9 site in the drought year (Fig. 4b).Thus the difference in suction in the surface soil in the drought summer was larger than in the non-drought year.In the sub-surface soil, there were no obvious differences in suction among the sites for both the drought and non-drought years.Cumulative transpiration of L chinensis and S. krylovii and soil water suction in drought and non-drought summers and average cumulative transpiration are shown in Figure 6.
The bar graphs show the cumulative transpiration at the end of July (DOY = 212).The cumulative transpiration of L. chinensis was slightly higher than that of S. krylovii in the nondrought year at all three sites.In contrast, the cumulative transpiration of S. krylovii was slightly higher than that of L. chinensis in the drought year at all three sites.Figure 7 shows that the suction of subsurface soils under growing S. krylovii was higher than that under L. chinensis in the drought year.Meanwhile, the soil suctions of surface and subsurface soils under the growing S. krylovii were higher than that under L. chinensis in the drought year.
structured in comparison with the SNG site.
It has been reported that the volume fraction of capillary pores for available water for plants and macro pores in an abandoned cropland was lower and higher, respectively, than those in grassland without a history of cultivation, because of destruction of the soil structure by past tillage [1].Soil water in abandoned fields, such as CA2, is retained in macropores at low suction.
Consequently, the suction in abandoned fields, especially in CA2, is lower than in SNG at a given volumetric water content.

Volumetric water content and suction of soil water in drought and non-drought summers
The lower content of coarse sand may lead to higher soil suction at the SNG site and at the later stage of abandoned fields, CA18, which may affect suction of the surface soil at each site after rainfall (Fig. 4a and 4b).The structured soil which has a lower coarse sand content and many capillary pores contributed to keeping the water at high suction.On the other hand, soil water at the CA2 site was retained in developing macropores consisting of coarse sand, whose size is too large to maintain a high capillary force.The differences in surface soil suction among the sites implie a difference in land-use history and periods of abandonment.The present study suggests that surface soil suction more clearly indicated soil hydraulic conditions for plant growth than the volumetric water content.Suction in the surface soils seems to indicate that the available water is restricted for drought tolerant plants at the CA18 and SNG sites and is less restricted at the CA2 site.
There were few differences in the volumetric water content and suction in the subsurface soils among sites except for the CA9 site.The high value of Wsat of CA9a and CA9c (Table 1) results in the differences in the volumetric water content and soil water suction observed at the CA9 site.The volumetric water content and suction in the subsurface soil were relatively more abundant (Fig. 3c and d) and lower (Fig. 4c and d) than in the surface soil, respectively.In general, the root depth of perennial species is deeper than for annual species, and the growth rate of perennial species relies on subsurface soil water.Therefore, the high suction of the surface soil at the CA18 and SNG sites has less possibility of preventing the growth of perennial species.
Although previous studies have shown changes in water retention curves and hydraulic conductivities of the near-saturated range of surface and subsurface soils [12,13] and of the unsaturated range (i.e., high suction range) of surface soils [1] due to crop abandonment, the present study indicates that suction of the surface soil water differed according to the period of abandonment and climatic conditions considering infiltration into and capillary rise from the subsoil.

Cumulative transpiration of L chinensis and S. krylovii and soil water suction in drought and non-drought summers
The higher cumulative transpiration of L. chinensis occurred in the days immediately following rainfall (e.g.Fig. 6a, DOY: 195) in the non-drought year, which may be caused by a difference in the reduction coefficient at around -1000 cm (Fig. 1).Leymus chinensis can maintain its potential transpiration rate up to a suction of about -1000 cm.This variation of cumulative transpiration along with the weather conditions shows why L. chinensis is the predominant species S. krylovii.Since annual species are dependent on the soil water in the surface layer, lack of available soil water in the surface is critical for them.These phenomena may indicate a competitive superiority of S. krylovii in dry Mongolian rangeland.
In abandoned cropland, the typical dry soil conditions were lost due to an increase in the fraction of coarse sand, especially at the CA2 site.The soil water is retained in the macropores at low suction (power), which means that this soil water could be used more easily than water at higher suction in the capillary pores, because soil water at low suction is preferred by annual plants.
In these abandoned croplands, typical perennial grasses, such as S. krylovii which has a tolerance for relatively high drought stress, would not have an advantage in this non-dry soil condition.In contrast, the increase in fine sand content of the surface soil layer with increasing period of abandonment and irregular drought would contribute to providing dry conditions and encourage the replacement of L. chinensis by S. krylovii.The present study suggests the importance of typical dry conditions of soil and climate for the growth of the typical perennial grass, S. krylovii, in degraded abandoned fields.This abiotic interaction between soil hydraulic properties and climate conditions appears to play an important role in plant succession in abandoned cropland.
Ecosystem Studies, Graduate School of Agricultural and Life Sciences, The University of Tokyo) for their helpful comments.This study was conducted under the Global Environmental Research

Table2 Parameter values of plant properties in theFig. 1 .
Fig. 1.Drought stress response function for L. chinensis and S. krylovii applied to the simulations of

Fig. 2 .
Fig. 2. Unsaturated hydraulic conductivity of soil samples at depths of 0-5 cm as a function of

Fig. 3 .
Fig. 3. Soil water-retention curves for soil samples at depths of 0-5 cm from the four sites:

Fig. 4 .Fig. 5 .
Fig. 4. Volumetric water content at depths of 2.5 and 12.5 cm at each site in non-

Fund
of Japan's Ministry of the Environment "Desertification Control and Restoration of Ecosystem Services in Grassland Regions of North-East Asia" (no.G-071), with the additional support of a JSPS Research Fellowship for Young Scientists awarded to A. Yanagawa (no.20-7002 and 24-7604).

Table 1
Parameter values of soil hydraulic properties in the hydrological simulation.134 Table1 Parameter values of soil hydraulic properties in the hydrological simulation.(continued) 135

Table 3
Particle size distribution of soils at the study sites.