Li-he Yin,Dn-dn Xu,Wu-hui Ji,Xin-xin Zhng,Jun Zhng

a Xi’an Center, China Geological Survey/Key Laboratory of Groundwater and Ecology in Arid Regions of China Geological Survey, Xi’an 710054, China

b Water Resources Research Institute of Shangdong Province, Jinan 250013, China

c Chinese Academy of Geological Sciences, Beijing 100037, China

Keywords:Groundwater depletion Phreatophytes Transpiration Numerical assessment Water table depth (DWT)Mean annual precipitation (MAP)(Hyper-) arid regions Hydrogeological survey engineering Northwest China.

ABSTRACT Quantitative assessment of the impact of groundwater depletion on phreatophytes in (hyper-) arid regions is key to sustainable groundwater management.However,a parsimonious model for predicting the response of phreatophytes to a decrease of the water table is lacking.A variable saturated flow model,HYDRUS-1D,was used to numerically assess the influences of depth to the water table (DWT) and mean annual precipitation (MAP) on transpiration of groundwater-dependent vegetation in (hyper-) arid regions of northwest China.An exponential relationship is found for the normalized transpiration (a ratio of transpiration at a certain DWT to transpiration at 1 m depth,Ta*) with increasing DWT,while a positive linear relationship is identified between Ta* and annual precipitation.Sensitivity analysis shows that the model is insensitive to parameters,such as saturated soil hydraulic conductivity and water stress parameters,indicated by an insignificant variation (less than 20% in most cases) under ± 50% changes of these parameters.Based on these two relationships,a universal model has been developed to predict the response of phreatophyte transpiration to groundwater drawdown for (hyper-) arid regions using MAP only.The estimated Ta* from the model is reasonable by comparing with published measured values.

1.Introduction

Water demands have been increasing significantly in recent decades due to the rapid development of agriculture and urbanization (Oki T and Kanae S,2006; Feng W et al.,2013; De Graaf I et al.,2014).Consequently,a continuous water table decline has been observed in many parts of the world (Gleeson T et al.,2012; Gorelick SM and Zheng C,2015; Pokhrel YN et al.,2015; Luong VV,2021; Boulariah O et al.,2021).This issue is particularly severe in (hyper-) arid regions where groundwater is the major water resource and has been over-exploited in many areas (Shakerkhatibi M et al.,2014; Jakeman A et al.,2016).As a result,phreatophytes(groundwater-dependent plants) in these regions are threatened by the increase of water table depth (DWT)induced by groundwater abstractions (Williams CA and Cooper DJ,2005; Palmer MA et al.,2008).DWT is often considered as a primary control on transpiration in areas with shallow water tables (Shah N et al.,2007; Wang P et al.,2014) which can be found in many parts of the world,even in deserts,due to the lateral convergence of groundwater (Fan Y et al.,2013).In these areas,groundwater provides a major water source to phreatophytes,particularly during dry periods.For example,poplars,groundwater-dependent vegetation,obtains 50% to 90% of the water from groundwater during dry periods (Snyder KA and Williams DG,2000).

Besides DWT,the response of phreatophytes to groundwater draw down also depends on mean annual precipitation (MAP),soil texture,physiological and morphological traits of phreatophytes (Mahoney JM and Rood SB,1992; Cooper DJ et al.,2003; Cooper DJ et al.,2006; Butler JJ et al.,2007).Previous studies have explored the physiological response of phreatophytes to groundwater level declines (Cooper DJ et al.,2003; Gazal RM et al.,2006)and as well as the impact of soil substrates (Mahoney JM and Rood SB,1992; Gonzalez E et al.,2010).As phreatophyte transpiration is supported by water from both saturated and unsaturated layers (Smith SD et al.,1998; Nippert JB et al.,2010; Danielescu S et al.,2020),transpiration is also strongly influenced by MAP that provides important water sources for plants via replenishing water in root zones (Shah N et al.,2007).It has been observed that transpiration is higher in humid areas or during post-rainfall periods when soil moisture is high (Zhang P et al.,2016).

With a declining water table,phreatophytes may suffer a certain degree of water stress.Unfortunately,a systematic assessment of the impact of falling water table on transpiration of groundwater-dependent plants in various climatic zones is not available so far.To the authors ’knowledge,limited case studies have been reported on the impact of water table decline on phreatophytesin (hyper-) arid regions (Table 1),mostly in the Great Basin,USA (Huntley D,1979; Nichols WD,1994; Cooper DJ et al.,2006; Elmore AJ et al.,2006; Groeneveld DP,2008; Pritchett D and Manning SJ,2012; Mata-Gonzalez R et al.,2012; Sanderson JS and Cooper DJ,2008),northwest China (Ma JX et al.,2013; Shen Q et al.,2015; Cheng DH et al.,2017; Liu B et al.,2017) and in Western Australia (FroendR and Sommer B,2010).In these case studies,evapotranspiration (ET),transpiration (T),or plant cover were used as indicators that decreased with an increase of DWT linearly (Mata-Gonzalez R et al.,2012; Pritchett D and Manning SJ,2012; Jayakody P et al.,2014; Shen Q et al.,2015; Liu ZY et al.,2016),exponentially (Nichols WD,1994; Cooper DJ et al.,2006;Elmore AJ et al.,2006; Sanderson JS and Cooper DJ,2008)or in a form of power or logarithmic functions (Groeneveld DP,2008; Ma JX et al.,2013).However,these relations are site-specific.A universal model describing the response of phreatophytes to falling water table across (hyper-) arid regions is not available.

About two-thirds of northwest China is characterized as(hyper-) arid zones,and many areas are suffering from groundwater depletion due to over-exploitation (Li CB et al.,2014; Liu X et al.,2015).The study areas in northwest China provide a range of climates to examine the impact of a falling water table on phreatophytes,suitable for developing a universal prediction model for (hyper-) arid regions.The objectives of this study are to (1) quantitatively explore the response of phreatophyte transpiration to water table drawdown by variable saturated flow modeling,(2)quantitatively assess the control of DWT and MAP on transpiration,and (3) develop a robust model to predict the impact of water table decline on phreatophyte transpiration over a range of MAP in (hyper-) arid regions.

2.Data and methods

2.1.Date sets

The broad study area covers northwest China with an area of about 3.3×106km2(Fig.1).Phreatophytes can be found in sites where the water table is shallow (usually 1-3 m),such as groundwater discharge areas and riparian zones (Wang P et al.,2014).The most common phreatophytes includePopulus euphraticaOliv.,Tamarix ramosissimaLedeb.,Alhagi sparsifolia,andSalix matsudana(Thomas FM et al.,2008;Yu TF et al.,2013; Wang P et al.,2014,Yin LH et al.,2014).Long-term meteorological data (1982-2006) for each site,including daily precipitation,wind speed,air temperature,relative humidity,and sunshine duration,were obtained from the China Meteorological Administration (http://data.cma.cn).MAP varies between 15 mm and 209 mm,and mean potential evapotranspiration (PET),calculated by the Penman-Monteith method,ranges from 933 mm/a to 1643 mm/a.To cover the full range of MAP,26 sites were selected (Fig.1; Table 2).The selection criterion for the 26 sites is that there is a station for every 10 mm increase in MAP unless there is not such a station available.

Table 1.Overview of field observations on the impact of water table decline on ET or T.

Soil profiles are about 10 km away from the meteorological stations as these stations are located in cities.Soil texture data were obtained fromthe China Soil Database(http://vdb3.soil.csdb.cn/) where the information of two layersis available,i.e.,0-0.3 m and 0.3-1.0 m.Pedotransfer functions were used to predict soil hydraulic parameters(Schaap MG et al.,2001).Leaf area index (LAI) around the soil profiles was obtained from MODIS LAI products(http://wist.echo.nasa.gov/) with 8-day 1 km resolution during 2000-2006.Before 2000 when MODIS LAI is unavailable,LAI values were obtained from GLASS LAI products(http://glass-product.bnu.edu.cn) with 8-day 5 km resolution during 1982-1999.Although 30m LAI during 1982-1999 canbe calculated using Landsat images,the lower temporal resolution (16 days) makes it difficult to capture the detailed temporal variations of LAI (Ganguly S et al.,2012).

2.2.Model setting

The HYDRUS-1D model (Simunek J et al.,2008) was used,which simulates subsurface water flow in variably saturated porous media using the Richards equation,and transpiration by a soil water potential-dependent stress function and potential transpiration determind by the atmospheric demand.A soil profile of 10 m was used for all sites as the maximum rooting depth in arid deserts can reach 9.5±2.4 m (Canadell J et al.,1996).Beyond this depth,the impact of groundwater onvegetation transpiration is very weak.The model domain was discretized with a constant interval of 0.01m.

The upper boundary was set as an atmospheric boundary condition,i.e.,daily rainfall and PET (Allen RG et al.,1998).PET was further partitioned into potential transpiration and potential evaporation using Beer’s law based on LAI (Ritchie JT,1972).Numerous field observations in northwest China suggest that vegetation coverage decreases with DWT at a certain range (5 m was used in this study),such as in the Horqin Sandy land (Duan L et al.,2011),in the Heihe River Basin (Wang P et al.,2011; Liu B et al.,2017) and the Tarim River basin (Xu HL et al.,2007).However,the decline rates are variable among sites.As a result,a linear decrease of LAI with different rates (i.e.,5%/m,10%/m,and 15%/m) was applied to investigate the effect of LAI on the results.In other words,a few scenarios were used in which LAI decreases by 20%,40%,and 60% with the decline of the water table from 1 m to 5 m,respectively.

The lower boundary was defined as a constant pressure head condition,with an initial depth at 1 m.The constantwater table assumption is valid at sites with deep water tables;however,water tables fluctuate temporally in response to climatic and surface water level variations as well as human activities when they are shallow (Novakowski KS and Gillham RW,1988; Horton JL et al.,2001; Liu ZY et al.,2016).To assess the uncertainty of constant lower boundary,a dynamic water table and a constant water table defined by the averaged dynamic water table (around 1 m) at three sites were compared,namely Tiekelike,Yinchuan,and Dulan (Fig.1;Table 2).The dynamic water table was generated by a model whose settings were the same as the above-mentioned model except that the lower boundary was defined as a prescribed upwards flux boundary,representing the input from lateral groundwater inflow.

Table 2.25-year MAP,25-year mean annual PET,soil texture,rooting depth,and climatic dryness zones (CDZ) including hyper-arid (HA) and arid (A) in the northwest China.

Fig.1.Location of the study sites in northwest China labeled in the order of increasing MAP; refer to Table 2 for site names.Each color represents a climatic zone,i.e.,hyper-arid (red) and arid (green).

In water-limited environments,the architecture of plant roots has been investigated through a global data set taken from the literature (Canadell J et al.,1996; Schenk HJ and Jackson RB,2002).However,only limited data are available for a specific region.Numerical and analytical modeling are the widely used methods to obtain representative root profiles(Collins DBG and Bras RL,2007; Guswa AJ,2008; Yang Y et al.,2016).In this study,a model developed by Gale MR and Grigal DF (1987) was used to generate root distribution as global analysis of root profiles indicates that this model was successful (Equation 1; Jackson RB et al.,1996).

where RD is the cumulative percentage of roots from the land surface to depth d,and γ is a rooting distribution index.Considering the initial DTW,the thickness of the root zone was assumed to be 1 m as field observations indicate that phreatophytes develop roots just above the water table(Naumburg E et al.,2005; Yin LH et al.,2015).

The S-shaped water stress function of root water uptake was used (van Genuchten MT,1987).Two parameters of this function,h50(a parameter at which transpiration is reduced by 50%) andp(an exponent parameter of the water stress response function),determine how transpiration is reduced from the potential value due to root-zone moisture stress.Typical values (Simunek J et al.,2008) were specified forh50(−800 cm) andp(3).

Salinity stress and oxygen stress to root water uptake were not considered.There are huge saline regions in northwest China due to high potential ET and flooding irrigation (Wang YG and Li Y,2013).Therefore,many native phreatophytes are salt-tolerant,such as the most popular phreatophytes in northwest China,Populus euphraticaOliv,andTamarix sinensisLour.They adapt to saline environments and can grow up well in heavy saline soils (Gu RS et al.,2004).Phreatophytes in arid regions usually do not develop roots in permanently saturated zones (Naumburg E et al.,2005).In response to water table falling or rising,roots distribution will be adjusted to maintain water uptake in unsaturated zones just above the water table to avoid oxygen stress (Jarrell WM and Virginia RA,1990; Oosterbaan A and Nabuurs GJ,1991;Segelquist CA et al,1993; Scott ML et al.,2000; Martin D and Chambers J,2002).

The soil texture within the top 1 m was defined using the available data and the soil texture from 1 m to 10 m was assumed to be the same as the lowest available soil types.The simulation was conducted from April,1stto October,31 of each year on a daily time step as this period is the growing season in northwest China.The initial conditions of the model were obtained from a spin-up model that used meteorological data of the first year (i.e.,the Year 1982) as the upper atmospheric boundary.As previous studies indicate that the spin-up time will increase for areas with deep water tables and dry climates (Rahman MM and Lu M,2015; Seck A et al.,2015),the model was run for 500 times (i.e.,500 years spinup) and the output of the final simulation was used as the initial condition.A 25-year simulation (1982-2006) was then implemented to assess the controls of DWT and MAP on the response of groundwater-dependent vegetation to the prescribed water table decline.The long-term simulation aims to obtain the mean response of phreatophytes to falling water tables.To explore the impact of DWT on phreatophyte response to groundwater drawdown,actual transpiration (Ta)was simulated under different DWT scenarios for each site.For a given DWT,the long-term (1982-2006) meanTawas calculated to avoid the impact of extreme climate in a specific year on the modeling results.

Even in a similar environment,the ET of phreatophytes is different due to various physiological processes,such as stomatal control.As indicated by the previous study(Steinwand AL et al.,2001),transpiration of three phreatophytes in the Owens Valley,USA was significantly different.Globally,rates of water use differ significantly among species,ranging from 10 kg/d to 1180 kg/d(Wullschleger SD et al.,1998).To make a comparison among sites and to develop a universal prediction model,Taat different water table depths was normalized by theTaat DWT =1 m (hereafterTa*) for each site.Phreatophytes in (hyper-)arid zones usually grow in riparian zones where water tables are very shallow (Sun ZY et al.,2015).Within the range of 0.3-1.0 m,transpiration is nearly constant due to the capillary rise (Shah N et al.,2007).It means that there will be less impact on transpiration following a water table decline within this range.Therefore,Ta*was defined at 1 m.

The simulations have uncertainty.For example,as only the top 1 m soil texture data is available,key information of soil hydraulic parameters may be missing,such as soil hydraulic parameters just above the water table which determine the height of capillary rise.A series of simulations were performed in which individual parameterswere altered and one parameter for the plant water stress function (i.e.,h50),while others were kept the same as their baseline values.The exception is the parameternthat was only increased by 50% as it should be higher than 1.0 in HYDRUS-1D.by ±50% for saturated hydraulic conductivity (Ks),α andn,

3.Results and discussion

The water table fluctuations in the selected three sites were shown in Fig.2,with mean DWT values of 1.04 m,1.01 m,and 1.08 m in Tiekelike (Fig.2a),Yinchuan (Fig.2b),and Dulan (Fig.2c),respectively.The water table rose due to the combined contributions from rainfall infiltration (data not shown) and lateral groundwater inflow,and declined as ET consumption was higher than lateral groundwater inflow.The simulated transpiration under the dynamic and constant water table conditions indicated that the differences were only 2.1%,1.3%,and 2.1% for the three sites,respectively.The reason might be that high water tables during wet periods or years benefited phreatophytes and low water tables during dry periods or years hurt phreatophytes.Considering these minimal differences and a meaningful long-term impact of water table decline on phreatophytes was interesting,the short-term dynamics were ignored and the constant water table at the low boundary was used.

3.1.Phreatophyte transpiration responses to water table depth

The simulatedTa*at three selected sites can be found in Fig.3.The three sites represented one of the driest sites (Fig.3a),the wettest site (Fig.3c),and a site in between (Fig.3b).When LAI reduction with water table decline was considered,Ta*decreased accordingly,and varied between sites (Fig.3).In Tiekelike,the decrease ofTa*was negligible (Fig.3a).In Yinchuan,the decrease ofTa*was 8.0%,16.8%,and 26.6%,with 20%,40%,and 60% of decrease in LAI,respectively(Fig.3b).In Dulan,the decrease ofTa*was similar to that in Yinchuan and reached 6.5%,14.4%,and 24.4%,respectively(Fig.3c).Considering the insignificant differences ofTa*relative to the LAI reduction rates,the results from the constant LAI were used to study the relationship,representing the optimistic response of phreatophytes to water table decline.

An exponential relationship was found betweenTa*and DWT (Fig.3),which was consistent with the results from previous observations or modeling studies (Nichols WD,1994; Cooper DJ et al.,2006; Elmore AJ et al.,2006;Sanderson JS and Cooper DJ,2008).When the water table was within the range of 1-4 m,Ta*declines dramatically following a small rate of groundwater level drop.However,Ta*kept almost constant when the water table was deeper than 4 m (Fig.3).The reason was that soil moisture at root zones was strongly influenced by groundwater when the water table was shallow (Chen X and Hu Q,2004; Soylu ME et al.,2011).Modeling studies indicated that when the water table was shallow,soil moisture was 21% higher than in the case where the water table was deep (Chen X and Hu Q,2004).The influence of groundwater on soil moisture generally occured within a depth range of 1 -5 m (Kollet SJ and Maxwell RM,2008).Consequently,water table decline would have a significant effect on transpiration when it was shallow.When the water table was below 5 m in depth,groundwater level drop would have a minimal effect on transpiration,indicated by a nearly constantTa*(Fig.3).

The rapid decline ofTa*indicated that water table drop would pronouncedly affect phreatophyte transpiration when the water table was shallow; even small declines at such sites could result in a significant change in transpiration as indicated in Table 1.For example,88% mortality of cottonwoods was observed following a water decline of 1.12 m (Scott ML et al.,1999).Therefore,phreatophytes were sensitive to water table variations in (hyper-) arid regions.For example,Euphrates poplar in the hyper-arid lower reaches of the Tarim River decreased by 87.6% following a water table decline caused by surface water diversion (Halik U et al.,2019).

Fig.2.Dynamic water table (solid lines) and the mean water table(dashed lines) for Tiekelike (a),Yinchuan (b),and Dulan (c).

Fig.3.Relation between the normalized Ta* and depth to water table for the selected sites.

3.2.Variation of phreatophyte transpiration with precipitation

To assess the control of precipitationon the response of phreatophyte transpiration to water table decline,the decrease was calculated inTaafter the water table dropped from DWT =1 m to DWT = 2 m and the range in which transpiration was most sensitive to groundwater level (Fig.3).The ratio ofTaat 2 m DWT toTaat 1 m DWT (hereafterTa1-2*) was used to indicate the response of vegetation to water table drawdown with different amounts of annual precipitation.A positive linear relationship betweenTa1-2*and annual precipitation was found (Fig.4),indicating that phreatophytes would suffer more severe water stress in dry years following a water table decline.For a similar groundwater level drawdown at a given site,only premature leaves of poplar trees became yellow in a humid period; in comparison,canopy dieback occurred in a dry period (Yin LH et al.,2018).Globally,over 88 cases of forest mortality were related to drought in southern Europe and western North America (Allen CD et al.,2010).

Fig.4 also suggested that the impact of groundwater drawdown on groundwater-dependent ecosystems is relatively more significant in dry sites than that in humid sites.Phreatophytes use both groundwater and soil moisture(Nippert JB et al.,2010; Dai Y et al.,2015).Soil water contents are higher in regions where MAP is higher as shown in the global distribution of surface soil moisture (McColl KA et al.,2017),therefore phreatophytes can rely on soil water when the water table falls.For example,the water sources ofTamarixswitched from primary dependence on groundwater to vadose-zone water after the water table declined (Nippert JB et al.,2010).It should be noted that the soil texture of each site,in addition to MAP,also influencesTa1-2*.As shown in Fig.4,Ta1-2*is about 0.48 in Yinchuan at 100 mm of annual precipitation (Fig.4b),while it is only 0.34 in Dulan at the same value of annual precipitation (Fig.4c).This difference is due to the content of fine material in soil being higher in Yinchuan than Dulan (Table 2).Previous studies indicate that plants will suffer more severe water stress in coarse substrates(Mahoney JM and Rood SB,1992; Bhattacharjee J et al.,2008).

3.3.Development of a prediction model

Based on the established relationships (Figs.3,4),an equation was proposed for predicting the impact of water table decline on the transpiration of phreatophytes:

whereTa*is the normalizedTa(-),aandbare coefficients(-),Pis the mean annual precipitation (m/a),andDis the depth to water table (m).The coefficients,aandb,were deduced from the modeling results (Fig.3).A series of the coefficients were calculated (Table 3) for the selected combinations of drawdown from 1 m to the maximum 5 m as the impact of the water table decline below 5 m is negligible(Fig.3).

The estimated coefficients from different drawdowns varied in a narrow range,commonly less than 20%.Thus,they were averaged for each site.A relationship between the coefficients and MAP was found (Fig.5).The deviations in Fig.5might be due to soil texture as the values ofTa*were different at sites with similar MAP,but different soil textures(Fig.4; Table 2).Usually,fine soils had greater water retention capacity and higher evaporation rates that would buffer the impact of water table decline on soil moisture(Moore RE,1939; Wösten JHM et al.,2001).Regression models were then constructed to relate the coefficients to MAP.Substituting the functions in Fig.5 into Equation(2)resulted in a universal predictive expression for quantifying the impacts of groundwater drawdown on transpiration across a range of MAP.

Fig.5.Dependence of Equation (1) coefficients a (a) and b (b) on MAP.

WhereTa*is the normalizedTa(-),Pis the mean annual precipitation (m/a),andDis the depth to water table (m).

3.4.Sensitivity analysis

Fig.4.Relation between the normalized Ta (the ratio of Ta at 1 m to Ta at 2 m) and annual precipitation for the selected sites.

The sensitivity ofTa*to different parameters on soil texture and plant water stress function varied (Table 4).However,the results indicated that the simulation results were not sensitive toKsandh50.Ta*varied slightly (2.1% to 17.9%)under ±50% changes of these parameters (Table 4).Ta*weren.As the variation ofTa*was insignificant (Table 4),a neglect not more sensitive to α and the most sensitive parameter was of the soil texture effect and the simplification of the plant water stress function during the development of the relationship between MAP andTa*was acceptable.It is known that transpiration in water-limited regions depends heavily on soil texture and water stress parameters.When a parameter changes,transpiration will systematically become higher or lower at all depths.Therefore,Ta*(a ratio of transpiration between different depths) has a lower variance than the absolute value of transpiration and is insensitive to these parameters listed in Table 4.

Table 3.Coefficients of a and b in Equation (2) for different cases at the selected sites＊.

Table 4.Sensitivity analysis of mean variation of Ta＊ to Ks,α,n,h50,and p.

3.5.Model validation

The performance of Equation(2) was tested against field measurements which cover a large range of water table variations (Fig.6).Fig.6a was based on sap flow of Populus measured by the compensation heat-pulse technique (Ma JX et al.,2013) and Fig.6b was based on plant cover or NDVI as previous studies suggest that ET was closely related to these parameters (Nagler PL et al.,2005; Glenn EP et al.,2010).The coefficients of determination (R2) were 0.99 and the root means square error (RMSE) varied between 0.01 and 0.03.In the Tarim Basin of China,a hyper-arid region with a MAP of less than 50 mm,the observed relationship between sap flow of populous and DWT was perfectly matched to the prediction by Equation(2) (Fig.6a).In the Owens Valley of USA,an arid region with a MAP of about 130 mm,vegetation depended on groundwater when DWT was less than 2.5 m(Elmore AJ et al.,2006),and a good agreement between the measured and predicted response to water table decline was observed (Fig.6b).

After the model validation,a contour map ofTa*was produced based on Equation(2) (Fig.7),from whichTa*could be obtained with a known MAP at sites with shallow water tables.The strongest response occured at the initial period of water table decline (1 -3 m),with a decrease by about 60%-90% depending on the MAP of each site.For example,at a site with 10 mm MAP,Ta*declines from 1.0 to 0.37 following a slight water table drop from 1.0 m to 1.1 m (Fig.7).The response becomes relatively stable with a water table below 3 m.

Fig.6.Comparison between predicted and measured Ta* in the Tarim Basin,northwest China (Ma et al.,2013) (a) and the Owens Valley,USA(Elmore et al.,2006) (b).

Fig.7.Contour diagram of Ta* based on Equation(3).

3.6.Limitations and further research

The prediction model is only applicable to (hyper-) aridregions with a shallow water table and may have a large bias in other climatic areas.Hydraulic redistribution (HR) and root water uptake compensation (RWUC) were ignored in the modeling.HR describes the passive movement of water amongsoil moisture at different depths via plant root systems(Prieto I et al.,2012; Gou S and Miller G,2014) and may play a significant role when the water table is deep or during dry periods (Miranda MT et al.,2018).Both numerical and experimental studies indicate that hydraulic redistribution increases transpiration to a certain level (Caldwell MM and Richards JH,1989; Brooks JR et al.,2002; Ryel R et al.,2002).For example,hydraulic redistribution could increase transpiration up to 20.5% in some days indicated by a simulation model (Ryel R et al.,2002).When the moisture stress occurs in the upper soil layers,the reduced water uptake is compensated by an enhanced uptake from the lower wetter zones (Yadav BK et al.,2009).However,root water uptake at all depths will systematically become higher than the original cases,which would reduce the impact of HR on the results.Root growth (RG) was also not considered in the modeling.However,the root may grow to reach the water table(Mahoney JM and Rood SB,1992),which would increase transpiration due to access to deeper moisture.Neglecting HR,RWUC and RG may bring uncertainty toTa*.In the future,more complex modeling including these processes is required to improve the understanding of the response of phreatophytes to falling water tables.

4.Conclusion

An exponential relationship is found between normalized transpiration and DWT,indicating that water table decline has significant impacts on phreatophytes when the water table is shallow.A positive linear function between the normalized transpiration and MAP is found,suggesting water table drawdown will have larger effects on phreatophytevegetation at dry sites.A model capable of predicting the response of groundwater-dependent plants to water table decline is proposed,indicating that the strongest response occurs at the initial period of water table decline (1-3 m) with a decrease by 60%-90%,and then the response becomes small and stable.Good agreement between prediction from the model and field observations indicates that the simple model provides a reliable tool for water resources management in(hyper-) arid regions to quantitatively assess the impact of falling water tables on phreatophytes.

CRediT authorship contribution statement

Li-he Yin conceived the presented idea and prepared the manuscript.Dan-dan Xu set up modles and run simulations.Wu-hui Jia,Xin-xin Zhang,and Jun Zhang processed simulated data and drew all the figures.All authors discussed the results and contributed to the final manuscript.

Declaration of competing interest

The authors declare no conflict of interest.

Acknowledgement

This research was funded by projects of the China Geological Survey (12120113104100 and DD20190351),National Natural Science Foundation of China (41877199),and Shaanxi Science and Technology Department (2019TD-040,2021ZDLSF05-01).The first author also appreciates the support of the Honor Power Foundation for visiting research at UNESCO-IHE.The authors would like to thank Neville Robinson and Hua-de Guan for the critical reading of the manuscript.The data that support the findings of this study are available from the corresponding author upon request.