国家节水灌溉杨凌工程技术研究中心

Agricultural Water Management 247 (2021) 106713

Research paper

Predicting agroecosystem responses to identify appropriate water-saving

management in arid irrigated regions with shallow groundwater: Realization on a regional scale

Lvyang Xiong a, b, Xu Xu a, b,*, Bernard Engel c, Yunwu Xiong a, b, Quanzhong Huang a, b,

Guanhua Huang a, b

a Chinese-Israeli International Center for Research and Training in Agriculture, China Agricultural University, Beijing 100083, PR China

b Center for Agricultural Water Research, College of Water Resources & Civil Engineering, China Agricultural University, Beijing 100083, PR China

c Agricultural and Biological Engineering, Purdue University, 225 South University Street, West Lafayette, IN 47907, USA

A R T I C L E I N F O

Handling Editor - Dr Z Xiying

Keywords:

Ago-ecohydrological modeling Agroecosystem processes Scenario analysis

Irrigated watershed

Water-saving management

A B S T R A C T

Scenario analysis is the basis of developing rational water management practices (WMPs) for watersheds. How to predict future hydrological responses on a regional-scale is still a challenge for modeling work in irrigated watersheds with shallow groundwater environments. Therefore, this paper presents an efficient realization of predicting regional agroecosystem responses and searching for appropriate WMPs, through using a water balance-based, semi-distributed hydrological model (SWAT-AG). The scenario case study is carried out in the Jiyuan Irrigation System located in the Hetao of upper Yellow River basin, based on the calibrated and validated modeling work in our previous companion paper. Eight scenarios of water-saving practices (WSPs) are proposed, with consideration for reducing irrigation depth and controlling initial groundwater depth. Then the coupled responses of agroecosystem processes to various WSPs are predicted for the case study region in 2012 and 2013, mainly related to the groundwater depth, root zone soil water and salinity, and crop yield/natural vegetation biomass. Based on the analysis for proposed scenarios, the 100% of present irrigation depth combined with increasing initial GWD by 50 cm are recommended as appropriate WSPs for dry years, and the 80% of present irrigation depth combined with increasing initial GWD by 100 cm are recommended for wet years, in order to maintain good environmental conditions for both crops and natural vegetation. In addition, results show that SWAT-AG could overcome the scale/function limitations of traditional soil/crop models and also avoid computational issues of numerical models. We further point out that the scenarios in reality will be more complicated and comprehensive in time and space, and thus the predictions should be updated accordingly. Overall, this case study fully presents the feasibility and practicality of using the SWAT-AG model to realize the scenario response analysis and water management decision-making on a region scale for irrigated watersheds with shallow groundwater environments.

1. Introduction

Excessive irrigation and induced secondary salinization have been common issues faced by many arid irrigated watersheds with shallow groundwater (e.g. in Northwest China, Pakistan and Northwest India, South Australia, and Northern Iran) (Doble et al., 2006; Kahlown et al., 2005; Noory et al., 2011; Qureshi et al., 2008; Ren et al., 2016; Singh et al., 2010). Water-saving practices (WSPs) are thus necessary for regulating the agroecosystem processes and improving the

soil-water-salt environments in such watersheds (Khan et al., 2009; Pereira et al., 2007). The adopted WSPs typically involve two aspects: promoting canal conveyance service and improving field irrigation. The application of WSPs could efficiently save irrigation water, meanwhile, it also causes a decline of groundwater table to alleviate the salinity problem. However, the over-application of WSPs may lead to excessive lowering of water table and possible drought risk, because shallow groundwater exerts strong effects on supporting crops and natural vegetation during the water deficit period (Nosetto et al., 2009; Ren

* Corresponding author at: Chinese-Israeli International Center for Research and Training in Agriculture, China Agricultural University, Beijing 100083, PR China.

E-mail addresses: xionglvyang@cau.edu.cn (L. Xiong), xushengwu@cau.edu.cn (X. Xu), engelb@purdue.edu (B. Engel), yxiong@cau.edu.cn (Y. Xiong), huangqzh@cau.edu.cn (Q. Huang), ghuang@cau.edu.cn (G. Huang).

https://doi.org/10.1016/j.agwat.2020.106713

Received 29 April 2020; Received in revised form 31 October 2020; Accepted 19 December 2020

Available online 2 January 2021

et al., 2016). Previous research demonstrated that inappropriate WSPs have resulted in negative impacts on agroecosystems and degradation of non-irrigated natural vegetation (Bai et al., 2017; Karimov et al., 2014; Porhemmat et al., 2018; Ren et al., 2017; Xu et al., 2011; Xue and Ren, 2017; Yang et al., 2012). Therefore, a thorough and accurate under- standing is needed of the agroecosystem responses to various WSPs for arid agricultural watersheds with shallow groundwater.

Simulation models are quantitative and powerful tools for scenario prediction and analysis, which is very helpful to understand the effects of water management practices (WMPs) on agroecosystem processes (Ahmadzadeh et al., 2016; Hanson et al., 2008; Toumi et al., 2016). These modeling cases for optimizing water management practices are mainly related to the soil/crop model (field scale or distributed regional scale), groundwater model and watershed hydrological model (Ahmadzadeh et al., 2016; Singh et al., 2006; Xu et al., 2019, 2011). Nevertheless, for scenario prediction in arid regions with shallow groundwater, there are some obvious deficiencies or limitations for these models. Typically, scenario analysis using soil/crop model is generally based on user-specified changes of water table depths (Pereira et al., 2007; Ren et al., 2018), because the model could not provide the predictions of regional (shallow) groundwater tables. This greatly re- stricts its actual applicability in scenario predictions for regional scale issues. Meanwhile, the application of groundwater modeling is more focused on groundwater dynamics and water table management. Espe- cially, two important terms (groundwater recharge/evaporation) are over-simplified or empirically treated in scenario simulation (Doble et al., 2006; Xu et al., 2010). They are actually closely related to vadose zone moisture conditions, while they are not reflected in previous pre- diction studies (Xu et al., 2011). Thus, the limited functions of groundwater model may largely affect its prediction accuracy and practical use in scenario analysis. Moreover, from the previous litera- ture, the WMP scenario predictions with watershed-scale hydrological models are rarely reported for irrigated watersheds with shallow groundwater. This is primarily due to ago-ecohydrological processes are often oversimplified or not considered for shallow water table areas (Xiong et al., 2019). Therefore, how to achieve a reasonable prediction

of WMP scenarios on a regional scale is necessary to be investigated

under arid and shallow water table environments.

Taking into account the considerations above, the paper aims at proposing and presenting a reasonable realization (with SWAT-AG) for predicting regional agroecosystem responses to WSPs in arid irrigated regions with shallow groundwater. SWAT-AG is a watershed-scale hy- drological model proposed in our companion paper (Xiong et al., 2019). It is specifically enhanced to be able to simulate the complicated agro- ecosystem processes in arid and shallow groundwater conditions, mainly including the interactions of the vadose zone water, ground- water, salt and plant growth. Moreover, SWAT-AG, as a balanced model, is more practical and efficient for regional issues compared with nu- merical models, as it could avoid the problems of model stability and computation burden. A case study using prediction modeling work is conducted in this work in a typical irrigated region of the arid upper Yellow River basin, on the basis of our previous calibrated/validated modeling case (Xiong et al., 2019). Various WSPs scenarios were considered involving reducing irrigation amount and increasing initial groundwater depth. The agroecosystem responses were then predicted and analyzed to identify appropriate WSPs that can promote crop pro- duction and maintain natural vegetation growth.

2. Materials and methods

2.1. SWAT-AG model

SWAT-AG is a physically-based, distributed hydrological model that especially includes enhanced capability for simulating regional agro- and eco-hydrological processes. It was proposed by Xiong et al. (2019) with significant modifications on the basis of the SWAT framework. The

original SWAT model was developed for predicting the impact of man- agement practices on hydrological processes and related processes in complex watersheds with varying slopes, soils and land use conditions (Neitsch et al., 2009). However, the SWAT model is far from sufficient when applied in agricultural watersheds with shallow water table en- vironments. The SWAT-AG enhances modeling capability for simulation in agricultural watersheds with shallow groundwater.

The SWAT-AG model inherits the capability (i.e. hydrological modeling and assessment) and the advantages (e.g. modeling robust- ness, effectiveness and comprehensiveness) of SWAT. It is based on a water balance-based mechanism and adopts the framework of subbasin and HRU (hydrological response unit) division. Thus, it is efficient and robust in handling complex regional modeling. Meanwhile, the SWAT- AG enhances the capability of regional agro-hydrological process simulation in watersheds with shallow groundwater. In SWAT-AG, the soil water balance module is improved to describe the effect of capillary rise on soil profile above shallow groundwater. The modules of soil salt balance and salt stress are developed for quantifying soil salt dynamics and the effects on crop/vegetation growth, which is not contained in the SWAT. A necessary modification is also made for groundwater model- ling by changing from the original HRU-based calculation to subbasin- based one. This is more reasonable in irrigated areas with shallow groundwater where lateral exchanges of groundwater are frequent. The effects of water management infrastructures are also quantified in groundwater modelling by calculating canal seepage and ditch drainage. In addition, some functions are incorporated into the SWAT-AG to consider the field practices of soil bunds and surface mulching. These make SWAT-AG able to capture the complex interactions among agro- ecosystem processes (e.g. plant growth, soil water/salt dynamic and groundwater depth fluctuation), especially for arid climatic and shallow groundwater environments. Additional description of SWAT-AG can be found in Xiong et al. (2019).

2.2. Case study area and modeling

The case study was built on the previously calibrated/validated modeling work of the Jiyuan Irrigation System (Jiyuan) in Hetao Irri- gation District (Hetao) of the upper Yellow River basin. Hetao is a typical arid irrigated watershed with shallow groundwater environments. Long-

term flood irrigation has resulted in shallow water depths (mostly varying at 0.5–3.0 m during the year) and secondary soil salinization. Jiyuan is located in the western region of Hetao (40◦ 45’–40◦ 52’N,

106◦59’–107◦ 07’E) (Fig. 1). It has an area of 8490 ha, where 60.7% and

30.3% are farmland and natural land, respectively. The topography is very flat, varying from 1036 to 1043 m above mean sea level (MSL).

Small elevation differences (0–50 cm) exist among different land covers.

The natural land is usually distributed in the lower elevation compared with nearby farmland, usually with shallower groundwater tables. The average annual precipitation is about 140 mm, and the mean annual

evaporation for a 20 cm pan reaches approximately 2000 mm. The mean annual temperature is about 7 ◦C, with monthly average of 10 and 24 ◦C in January and July, respectively (Ren et al., 2016).

The irrigation and drainage system in Jiyuan primarily consists of:

(1) two main canals which allocate irrigation water from south to north through tributary canals; and (2) two main drainage ditches located in the middle part and on the east boundary, and discontinuous ditches in the west part (Fig. 1). Major crops in Jiyuan include sunflower and maize, taking up nearly 90% of farmland area in recent years. The natural vegetation is mainly shrubs and sparse grass, which is affected with varying degrees of salinization. The tamarisk (Tamarix chinensis Lour.) is the main species of shrubs according to the on-site survey. The land use of Jiyuan changed little while the crop pattern was adjusted according to the willingness of local farmers. The land use maps for 2012 and 2013 are presented in Fig. 2. The land use map was obtained from the remote sensing result of Landsat satellite images, while crop pattern was generated from the manual visual interpretation of Google Earth

Fig. 1. Location of the case study area and the observations (Hetao: Hetao Irrigation District; Jiyuan: Jiyuan Irrigation System; YCA: an experiment site called Yangchang canal command area) (drawn from Xiong et al., 2019).

Fig. 2. Maps of subbasins division (a) and land use in 2012 (b) and 2013 (c) (modified from Xiong et al., 2019).

images (about 0.5 m resolution).

The crops in Jiyuan are irrigated by surface basin irrigation with water from the Yellow River, including six water diversions during the crop growing season (from May to September), and the seventh diver- sion (called autumn irrigation) after crop harvest (Ren et al., 2016). The autumn irrigation is applied for salt leaching and water storage, allow- ing it to support crop growth in the early stage of the next year. Table 1 shows the recorded cropping and irrigation schedules for a wet year (2012) and a dry year (2013), respectively.

On the basis of the two-year observation experiments and multi- source data collection, the SWAT-AG modeling work for Jiyuan was carried out (Xiong et al., 2019). Jiyuan was divided into 28 subbasins based on the irrigation and drainage system in SWAT-AG (Fig. 2); and the subbasins are further subdivided into 290 HRUs and 350 HRUs in 2012 and 2013, respectively, through the combination of land use, crop

pattern, soil type and slope.

2.3. Model calibration and validation

The SWAT-AG model has been calibrated and validated in Jiyuan based on the observed data in 2013 and 2012, respectively (Xiong et al.,

2019). The observations included soil water storage and salinity con- centration (0–100 cm depth), leaf area index (LAI) and groundwater depth (GWD). They were collected from 28 sampling points and 39

groundwater monitoring wells in Jiyuan (Fig. 1). The model calibration was first conducted on the basis of detailed canal-scale data in YCA (Fig. 1), using SUFI2 algorithm in SAWT-CUP (SWAT Calibration Un- certainty Procedures) (Abbaspour et al., 2004). Regional data sets in Jiyuan were then used to further calibrate the model by trial-and-error method. Meanwhile, the model performance was also evaluated by

Present management schedules of Jiyuan in 2012 and 2013, modified from Xiong et al. (2019).

Maize

May 31 Watermelon June 3 Watermelon

Irrigation May 2–7 Sunflower 187 May 10–14 Sunflower 203

Wheat 135 Wheat 80

Watermelon 187 Watermelon 189

May 23–27 Wheat 76 May 23–27 Wheat 78

June 22–26 Maize 117 June 25–30 Maize 112

Sunflower 90 Sunflower 103

Wheat 76 Wheat 74

August 2–4 Maize 114 July 15–19 Maize 94

Sunflower 92 Sunflower 106

August 27–30 Maize 75 Aug. 6–10 Maize 108

Sunflower 90

Autumn irrigation October 10–15 Farmland 180 Oct. 26–31 Farmland 180

Harvest July 15 Wheat July 15 Wheat August 30 Watermelon August 30 Watermelon

September 20 Maize September 20 Maize

September 20 Sunflower September 20 Sunflower

comparing simulated results with remote sensing ETa (Yang et al., 2012) and the results in our previous studies (Ren et al., 2016, 2017, 2018). The calibrated values of sensitive parameters are presented in Table 2. The calibrated SWAT-AG well captured the soil water-salt dynamics, groundwater level fluctuations and crop growth process. The simulation of ETa was also reasonable compared with the remote sensing ETa in croplands and natural lands. The fitness indicators for model perfor- mance were all in acceptable ranges as compared with the canal-scale and region-scale observations (Table 3). A more detailed description about model calibration/validation could be found in Xiong et al.

(2019).

2.4. Using the model for predicting the WSPs scenarios

In order to show the advantages of SWAT-AG in prediction

Table 2

capability, the WSPs scenarios proposed involve the adoption of water- saving irrigation and changes of initial GWDs (Table 4). The present and two reduced irrigation strategies (i.e. 80% and 60% of present irrigation amount) are considered in scenario creation, noted as IR1.0, IR0.8 and IR0.6, respectively. Improving the irrigation and drainage system may lead to the lowering of the present water tables. Note that the initial GWD may have a significant effect on the subsequent hydrological dy- namics processes. Thus, the present and two increased initial GWD strategies (i.e. increased by 50 cm and 100 cm) are set in scenarios, noted as GWD0, GWD50 and GWD100, respectively. Summarizing, a total of nine scenarios (i.e. three initial GWDs combining three irrigation strategies, Table 4) are proposed to predict the possible agroecosystem responses in Jiyuan during the plant growing season (from May to September). Scenario S1 represents the present conditions. The climatic data in scenarios were set the same as in 2012 (wet year) and 2013 (dry

Sensitive parameters and their calibrated values for SWAT-AG (drawn from Xiong et al., 2019).

Process Parameter Description Rangea Calibrated valueb

Farmland Natural

land

Soil water AWC1 (0–30 cm) Available water content (cm3 cm—3) -0.2–0.2 0.27 0.28

SOL_K1 (0–30 cm) AWC2 (70–100 cm) SOL_K2 (70–100 cm)

Saturated hydraulic conductivity (mm hr—1) -0.2–0.2 5 3

Available water content (cm3 cm—3) -0.2–0.2 0.27 0.28

Saturated hydraulic conductivity (mm hr—1) -0.2–0.2 6.5 5

Farmland Natural

land

Groundwater Sy Specific yield of the shallow aquifer (-) -0.1–0.1 0.038 0.038

δgw Delay time for water leaves the soil profile to reach groundwater (days) – 1 1

Maize Sunflower Watermelon Natural

land

Crop growth BLAI Potential maximum LAI (-) – 5 5 4 1

DLAI Fraction of plant heat unit when LAI begins to decline (-) – 0.8 0.65 0.8 0.99

T_BASE Minimum temperature for plant growth (℃) – 8 6 18 12

PHU Potential heat units required for maturity of crop (℃) – 2000 2100 400 1800

br The reduction factor in ECe (%/(dS m—1)) -0.1–0.1 6 6 4.5 3

which salt stress happens (dS m—1)

Note:

a represents relative changes of parameters based on initial values for YCA and Jiyuan;

b represents calibrated values of parameters in YCA.

Table 3

The values of fitness indicators for simulations evaluated with the observation data (reorganized from Xiong et al., 2019).

YCA (Canal-scale) Jiyuan (Region-scale)

Land use Items NSE R2 RSR Items NSE R2 RSR

Calibration Maize Soil water storage 0.60 0.77 0.63 Soil water storage 0.62 0.72 0.61

(2013) Soil salt concentration 0.38 0.65 0.79 Soil salt concentration 0.87 0.87 0.37

Leaf area index 0.72 0.83 0.53 Groundwater depth 0.59 0.66 0.64

Sunflower Soil water storage 0.83 0.87 0.42

Soil salt concentration 0.30 0.52 0.84

Watermelon Soil water storage 0.57 0.87 0.66

Soil salt concentration 0.47 0.65 0.43

Natural land Soil water storage 0.48 0.65 0.72

Soil salt concentration 0.43 0.47 0.75

– Groundwater depth 0.85 0.89 0.39

Validation Maize Soil water storage 0.42 0.45 0.76 Soil water storage 0.61 0.62 0.63

(2012) Soil salt concentration 0.61 0.65 0.62 Soil salt concentration 0.89 0.89 0.33

Leaf area index 0.86 0.89 0.37 Groundwater depth 0.60 0.62 0.63

Sunflower Soil water storage 0.69 0.71 0.56 ETa 0.07 0.03 1.22

Soil salt concentration 0.56 0.65 0.66

Watermelon Soil water storage 0.74 0.81 0.51

Soil salt concentration 0.74 0.82 0.47

Natural land Soil water storage 0.22 0.23 0.88

Soil salt concentration 0.19 0.29 0.90

– Groundwater depth 0.81 0.85 0.44

Note: NSE is the Nash and Sutcliffe model efficiency, R2 is the coefficient of determination, RSR is the RMSE-observations standard deviation ratio, and their equations can refer to Xiong et al. (2019). ETa is the actual evapotranspiration.

Table 4

Scenarios of water-saving practices for Jiyuan. S1 refers to the baseline scenario.

3.
Results and discussion

3.1. Responses of GWD dynamics to various WSPs scenarios

Scenarios Reducing

irrigation depth

Increasing initial GWD

Scenarios in detail

The average GWD during the simulation period (AGWDsp) at the

S1 100% – IR1.0: the existing irrigation

S2 80% – IR0.8: reduce the irrigation

subbasin scale was calculated for various scenarios, as presented in Fig. 3. In Hetao, classes of GWD could be defined as: very-shallow

S3 60% –

depth to 80%

IR0.6: reduce the irrigation depth to 60%

(<1.0 m), shallow (1.0–1.5 m), reasonable (1.5–2.0 m), deep (2.0–2.5 m) and very-deep (>2.5 m), according to previous studies (Xu

S4 100% 50 cm GWD50: the initial GWD

et al., 2013; Ren et al., 2017, 2018). Our results showed that reducing

S5 80% 50 cm

S6 60% 50 cm

S7 100% 100 cm

increases 50 cm IR0.8 + GWD50 IR0.6 + GWD50

GWD100: the initial GWD increases 100 cm

irrigation depth and increasing initial GWDs resulted in the lowering of

water tables but only to a limited degree. The AGWDsp was not increased significantly for scenarios S1–5 and S7, mostly varying between 1.0 and

1.5 m. For the combination scenarios of S6 and S8, the AGWDsp

S8 80% 100 cm IR0.8 + GWD100

S9 60% 100 cm IR0.6 + GWD100

year). The changes in GWD conditions, soil moisture and salt balance terms (in the 0–100 cm soil profile), and crop yield/natural vegetation biomass were comprehensively evaluated for the case study area.

increased markedly and fluctuated in the reasonable range of 1.5–2.0 m

for many subbasins. However, the AGWDsp was deeper than 2.0 m in some subbasins for scenario S9 that is the most intensive WSPs.

When looking further at the spatial distribution of AGWDsp (i.e. on HRU scale), results demonstrated that WSPs brought both positive and negative responses in aspects of GWD changes. The spatial distributions of AGWDsp for scenarios S1, S5 and S9 in 2013 are selected and provided in Fig. 4a. Results showed that the GWDs were very-shallow or shallow in most HRUs for the baseline scenario S1. Additionally, for the mod- erate scenario S5, fewer HRUs had very-shallow GWD but there were a

Fig. 3. Simulated results of average groundwater depth during simulation period (AGWDsp) for various WSPs scenarios (on a subbasin scale): 2012 (a) and 2013 (b).

Fig. 4. Spatial distribution of average groundwater depth (AGWDsp) (a), soil water content (ASWCsp) (b) and electrical conductivity of saturated soil paste extract (ASECsp) (c) under different scenarios (S1, S5 and S9) in year 2013.

number of HRUs with shallow and reasonable GWD. However, the deep and very-deep GWD were widely distributed in scenario S9, which was mutually supported with the GWD changes at the subbasin scale (Fig. 3). Therefore, it implied that there were some risky responses of GWDs to the very intensive WSPs, especially in dry years.

3.2. Responses of soil water and salt conditions to various WSPs scenarios

3.2.1. Soil water

The average soil water content of the root zone during the simulation period (ASWCsp) was calculated at a subbasin scale, as shown in Fig. 5. Results showed that the ASWCsp were both decreased when reducing irrigation depth or increasing initial GWD. It decreased to a limited degree in the wet year (2012), and maintained relatively high levels

with 0.3–0.35 mm/mm and 0.35–0.40 mm/mm for farmland and nat-

ural land, respectively. By comparison, the ASWCsp showed a more

pronounced decline in the dry year (2013), despite that the ASWCsp is still around 0.30–0.35 mm/mm for some mild/moderate WSPs sce- narios (e.g. S2–5 and S7). For the intensive WSPs scenarios (e.g. S6, S8

and S9), ASWCsp has dropped lower than 0.30 mm/mm both for farm- land and natural land in many subbasins. Therefore, in terms of soil moisture, these responses indicated that most WSPs scenarios were acceptable for the wet year; however, some intensive scenarios may result in drought risk in some cases during dry years.

To further illustrate the soil water response to WSPs in the dry year, the spatial distribution of ASWCsp (on HRU scale) for scenarios S1, S5 and S9 in 2013 are selected and provided in Fig. 4b. The results of scenario S5 confirmed that ASWCsp maintained a relatively high level with moderate WSPs, meanwhile the spatial distribution of ASWCsp was not dissimilar to that in the baseline scenario (S1). This indicated that the moderate WSPs strategies should be acceptable in the dry year. However, the ASWCsp was dramatically decreased in scenario S9 (Fig. 4b). There were a number of HRUs having relatively lower ASWCsp, which demonstrated that intensive WSPs would cause drought risk during the dry year.

3.2.2. Soil salt

The average soil salt content of the root zone during the simulation period (ASSCsp) was calculated at a subbasin scale, as shown in Fig. 6. The ASSCsp for farmland increased with reduced irrigation, contrarily it decreased with increasing initial GWD. Therefore, the ASSCsp for farmland was decreased with increasing initial GWD alone (i.e. S4 and S7). Meanwhile, when combined WSPs were applied, the increase of ASSCsp for farmland was relatively insignificant in some scenarios (i.e.

S5 and S8–9), with less than 0.15 g/kg. However, note that the ASSCsp

for farmland could still be increased significantly with IR0.6 strategies (i.e. S3 and S6), with the increase of over 0.2 g/kg. This indicated that

scenarios S3 and S6 may result in severe salinity problems for farmland. Moreover, results showed that the ASSCsp change for natural land was insignificant in WSPs scenarios (Fig. 6).

The average electrical conductivity of saturated soil paste extract for the root zone during the simulation period (i.e. ASECsp) was further calculated. Its spatial distribution (on HRU scale) for the selected sce- narios S1, S5 and S9 is provided in Fig. 4c. The results of scenarios S5 and S9 were not dissimilar to that of the baseline scenario (S1), con- firming that the proposed WSPs strategies may have insignificant in- fluences on the salinity conditions in the root zone. Just a few HRUs underwent a significant increase of ASECsp in scenarios S5 and S9.

3.3. Responses of plant growth to various WSPs scenarios

3.3.1. Crop yield

The relative change of crop yield (ΔRY) was calculated for each crop type (on the subbasin scale), as shown in Fig. 7. The crop yield showed similar responses to WSPs in wet year (2012) and dry year (2013). Re-

sults showed that crop yield decreased with reduced irrigation depth alone. In S2 and S3, IR0.8 and IR0.6 could result in about 5–8% and 11–14% reduction of crop yield in 2012 and 2013, respectively. Similar

responses were also obtained for scenarios S4-S6 or S7-S9. Note that the significant decrease of crop yield took place in some subbasins where salinity was the critical stress factor, due to the severe root zone salinity problem with adopting the IR0.6 strategy. In contrast, the crop yield was

slightly increased (with ΔRY not greater than 3%) when only lowering

the initial water tables, e.g. according to the comparisons of S1, S4 and S7 (or S3, S6 and S9) in Fig. 7. It was found that the combined strategies could not only save irrigation water but also maintain acceptable yields,

e.g. scenario S8 with IR0.8 and GWD100.

The water-salt stress during the crop growth period was further calculated, with lower values indicating larger stress conditions. The results of maize growth for scenarios S1, S5 and S9 were selected and provided in Fig. 8. The water-salt stress was gradually alleviated for scenarios S1, S5 and S9 for the initial period (May and early June), when no irrigation was applied for maize. This meant that increasing initial GWD was beneficial to alleviate the root zone salinity problem during the initial season. By comparison, for the period after June, water-salt stress gradually became more severe with changing scenarios S1 to S5 and S9, which implied the negative impacts from the reduced irrigation. Moreover, we reason that the salinity problem was the critical factor for some cases of significant crop yield decrease under WSPs scenarios.

Fig. 9 plots the spatial results of ΔRY for HRUs under scenarios S5 and S9

in the dry year of 2013. By comparing the spatial results of ASECsp (Fig. 4), it was evident that the HRUs with low ΔRY were identical to the HRUs with significant increase of ASECsp.

Fig. 5. Simulated results of average soil water content during simulation period (ASWCsp) in the root zone for various WSPs scenarios (on a subbasin scale): 2012 (a) and 2013 (b).

Fig. 6. Simulated results of average soil salt content in the root zone during simulation period (ASSCsp) for various WSPs scenarios (on a subbasin scale): 2012 (a) and 2013 (b).

Fig. 7. Predicted average values of relative change of yield (ΔRY) and relative change of biomass (ΔRB) for various WSPs scenarios (on a subbasin scale): 2012 (a) and 2013 (b).

3.3.2. Natural vegetation biomass

Fig. 7 presents the results of relative change of natural vegetation biomass (ΔRB) for subbasins under WSPs scenarios in 2012 and 2013, respectively. The natural vegetation biomass had different responses to

WSPs between the wet year (2012) and the dry year (2013). In the wet year, natural vegetation biomass could be maintained and even slightly

increased for most WSPs scenarios (with ΔRB<5%). Only a slight

decrease was observed in very intensive scenario S9. In addition, the water-salt stress during the growth period is presented in Fig. 8, taking S1, S5, and S9 as examples. The stress value under scenario S5 was a little higher than that for scenario S1, with both of them higher than that in scenario S9. This further shows WSPs may not affect natural vegeta- tion growth in the wet year. However, in contrast, natural vegetation biomass mainly showed a decreased trend in the dry year, while only

increasing in some subbasins for mild scenarios (i.e. S4 and S5). The ΔRB

values reached ranges between 10% and 15% for some scenarios (e.

g. S6, S8 and S9). We also noted that the water-salt stress became more severe in S9 compared with S1 and S5 (Fig. 8), due to less groundwater contribution. The analysis indicates severe negative effects on natural vegetation growth may occur for some proposed WSPs during dry years. In addition, the responses of natural vegetation biomass are strik- ingly different in space in the dry year, which was attributed to the

different responses of soil moisture conditions. Fig. 9 plots the spatial distribution of ΔRB (on HRU scale) in scenarios S5 and S9 in the dry year (2013). We observed the HRUs that had a significant decrease of natural

biomass (Fig. 9) were mainly identical to those with relatively lower ASWCsp (but not with higher ASECsp) (Fig. 4). This implied that the change of moisture conditions (rather than salt conditions) should be the

critical factor to cause the natural biomass decrease in WSPs scenarios.

3.4. Discussion

Our predictions provided comprehensive agroecosystem responses to the proposed WSPs in the case of Jiyuan. Results indicated that several scenarios (S4, S5 and S7 and S8) were feasible to maintain relatively good environments of soil water and groundwater for plant growth during the wet year. Other scenarios (S2, S3, S6 and S9) may result in some different agroecosystem problems and the reduction of crop yields. S3 and S6 mainly caused a significant increase of soil salinity, while S9 brought the risk of drought problem for plant growth. Furthermore, in the above-mentioned feasible scenarios, the scenario of S8 (IR0.8 GWD100) can not only reduce irrigation water but also maintain acceptable yield and natural vegetation biomass, which could be recommended as the appropriate WSPs for the wet years. However, during dry years, the scenarios related to reducing irrigation depth would all result in significant reduction for both crop yield and natural vegetation biomass. Only scenario S4 (IR1.0 GWD50) could maintain good growth both for crops and natural vegetation. Therefore, scenarios S4 and S8 are suggested as appropriate WSPs in Jiyuan for dry years and wet years, respectively, compared with other scenarios. Moreover, the response analyses showed that the averaged salt content in the root zone had an increasing trend when reducing the current irrigation water amount. This implied that it was necessary to increase extra fraction of irrigation water for leaching salts during crop growing in some local areas. Some other practical agronomic measures can also be adopted to alleviate the salinity effects, such as the ridge tillage, surface residue

Fig. 8. Combined water and salt stress on maize and natural vegetation for the selected WSPs scenarios (S1, S5 and S9), during simulation period for the year 2012

(a) and 2013 (b).

Fig. 9. Spatial distribution of relative change of yield (ΔRY) (a) and relative change of biomass (ΔRB) (b) in year 2013.

cover and organic fertilizer application. In addition, another flood irri- gation event is often applied to leach salt after harvest in irrigation districts of upper Yellow River basin (e.g. between October and November in Hetao). This is efficient to reduce the accumulated salt during crop growth season, and to prevent the inter-annual salt accu- mulation in the long term.

In this study, we used a group of case studies to search for the appropriate WSPs using SWAT-AG. The scenario analysis with SWAT-AG can provide a full understanding of regional agroecosystem responses in irrigated watersheds with shallow groundwater, which could

significantly outperform and broaden the previous research in scenario prediction to the regional scale. Compared with studies of Pereira et al. (2007) and Ren et al. (2018) that relied on user-specified (presumptive) GWDs, our study captured groundwater responses and thus allowed us to predict regional responses to WSPs directly without specifying groundwater conditions. Compared with studies of Doble et al. (2006) and Xu et al. (2010) that oversimplified groundwater recharge/evapo- ration terms, our study simulated vadose zone water dynamics and thus predicted the groundwater responses from a realistic perspective.

In fact, practical regional water-saving management could be much

more complex, because the management varied in both space and time. Even through the overall responses were positive, there were some poor soil water conditions and decrease of crop yield and natural vegetation biomass in some local areas. When designing the real WSPs, more factors and responses should be considered for increasing the water-use effi- ciency and economic benefits. The real application should be based on more realistic water-saving scenarios, taking the spatial variation of management into consideration (e.g., crop pattern adjustment and spatially different irrigation schedules). Moreover, SWAT-AG can also be applied for timely scenario simulation and water management with integration of big data technology in investigations that follow.

4. Conclusions

This study provided an efficient realization of scenario prediction on a regional scale and identification of appropriate water-saving practices (WSPs) for arid irrigated watersheds, using a water balance-based, semi- distributed hydrological model. The SWAT-AG model proposed in our previous companion paper was employed to predict WSPs scenarios, considering strategies of reducing irrigation depth and controlling initial groundwater depth (GWD). The scenario case study was conducted based on the previous calibrated/validated modeling work in the Jiyuan Irrigation System located in the Hetao of the upper Yellow River basin. Results indicated that scenario prediction with SWAT-AG could provide a full understanding of regional agroecosystem responses, involve GWD, and soil water-salt conditions in root zone and crop/natural vegetation growth.

Results showed that reducing irrigation depth and increasing initial GWD both caused the lowering of water tables but only to a limited degree. AGWDsp would not exceed 2.0 m for WSPs scenarios other than

the most intensive scenario S9 (IR0.6 + GWD100). Although the WSPs

resulted in a decrease of soil moisture in the root zone, the ASWCsp was

still relatively high during the wet year; while it can maintain an acceptable level (> 0.3 mm/mm) for mild/moderate WSPs scenarios (e.

g. S2–5 and S7), but may cause drought risk for some intensive scenarios

(e.g. S9) during the dry year. Meanwhile, the changes of root zone salinity for farmland was insignificant (less than 0.15 g/kg) for most WSPs scenarios, with only a marked increase for the IR0.6 strategy. There were also insignificant changes of root zone salinity for natural land. Correspondingly, the crop yield showed similar responses during the wet year and dry year, i.e., decreased with reducing irrigation while

slightly increased with increasing initial GWD. The IR0.6 strategies could result in 11–14% reduction of crop yields. Natural vegetation biomass could maintain and even slightly increase for most WSPs sce- narios in wet years; however, it had a 10–15% reduction in some intensive scenarios during the dry year.

On the basis of scenario analysis, the 100% of present irrigation depth combined with increasing initial GWD by 50 cm are recom- mended as appropriate WSPs for dry years, and the 80% of present irrigation depth combined with increasing initial GWD by 100 cm are recommended for wet years, in order to reduce irrigation water and maintain crop yield/plant biomass. In addition, there were also negative effects occurring in some local areas, due to the spatial variance of agroecosystem responses. More comprehensive and complicated stra- tegies should be considered in realistic scenario predictions. Moreover, the simulation case studies showed that SWAT-AG could overcome the scale/function limitations of traditional soil/crop models, and avoid computation issues of numerical models in scenario prediction. Overall, our study efficiently increased the usability of the simulation model to support regional water management in arid irrigated watersheds with shallow groundwater.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence

the work reported in this paper.

Acknowledgements

This research was supported by the National Natural Science Foun- dation of China (grant numbers: 51639009, 51679235 and 52022108) and the 13th Five-year National Key Research and Development Pro- gram of the Chinese Ministry of Science and Technology (grant numbers: 2017YFC0403301).

References

Abbaspour, K.C., Johnson, C.A., van Genuchten, M.T., 2004. Estimating uncertain flow

and transport parameters using a sequential uncertainty fitting procedure. Vadose Zone J. 3 (4), 1340–1352.

Ahmadzadeh, H., Morid, S., Delavar, M., Srinivasan, R., 2016. Using the SWAT model to assess the impacts of changing irrigation from surface to pressurized systems on water productivity and water saving in the Zarrineh Rud catchment. Agric. Water

Manag. 175, 15–28.

Bai, L., Cai, J., Liu, Y., Chen, H., Zhang, B., Huang, L., 2017. Responses of field

evapotranspiration to the changes of cropping pattern and groundwater depth in large irrigation district of Yellow River basin. Agric. Water Manag. 188, 1–11.

Doble, R., Simmons, C., Jolly, I., Walker, G., 2006. Spatial relationships between vegetation cover and irrigation-induced groundwater discharge on a semi-arid

floodplain, Australia. J. Hydrol. 329, 75–97.

Hanson, B.R., Hopmans, J.W., Simůnek, J., 2008. Leaching with subsurface drip irrigation under saline, shallow groundwater conditions. Vadose Zone J. 7, 810–818.

Kahlown, M.A., Ashraf, M., Zia-ul-Haq, 2005. Effect of shallow groundwater table on crop water requirements and crop yields. Agric. Water Manag. 76, 24–35.

Karimov, A.K., Sˇimůnek, J., Hanjra, M.A., Avliyakulov, M., Forkutsa, I., 2014. Effects of

the shallow water table on water use of winter wheat and ecosystem health: Implications for unlocking the potential of groundwater in the Fergana Valley

(Central Asia). Agric. Water Manag. 131, 57–69.

Khan, S., Hanjra, M.A., Mu, J., 2009. Water management and crop production for food

security in China: a review. Agric. Water Manag. 96, 349–360.

Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., 2009. Soil and water assessment tool theoretical documentation version 2009, Texas Water Resources Institute.