Article Request Page ASABE Journal Article
Optimizing Irrigation Rates for Cotton Production in an Extremely Arid Area Using RZWQM2-Simulated Water Stress
C. Liu, Z. Qi, Z. Gu, D. Gui, F. Zeng
Published in Transactions of the ASABE 60(6): 2041-2052 (doi: 10.13031/trans.12365). Copyright 2017 American Society of Agricultural and Biological Engineers.
Submitted for review in March 2017 as manuscript number NRES 12365; approved for publication as part of the “Crop Modeling and Decision Support for Optimizing Use of Limited Water” collection by the Natural Resources & Environmental Systems Community of ASABE in September 2017.
The authors are Che Liu, Graduate Student, and Zhiming Qi,ASABE Member, Assistant Professor, Department of Bioresource Engineering, McGill University, Sainte-Anne-de-Bellevue, Quebec, Canada; Zhe Gu, ASABE Member, Graduate Student, Research Center of Fluid Machinery Engineering and Technology, Jiangsu University, Zhenjiang, China; Dongwei Gui, Associate Research Scientist, and Fanjiang Zeng, Research Scientist, Cele National Station of Observation and Research for Desert Grassland Ecosystems, Chinese Academy of Sciences, Urumqi, China. Corresponding author: Zhiming Qi, 21,111 Lakeshore Road, Sainte-Anne-de-Bellevue, QC, Canada H9X 3V9; phone: 514-398-7983; e-mail: email@example.com.
Abstract. Quantifying crop water demand and optimizing irrigation management practices are essential to water resource management in arid desert oases. Agricultural systems modeling can serve to develop a better understanding of the hydrologic cycle under various irrigation and climate conditions. RZWQM2-simulated water stress can be used as an indicator for irrigation scheduling but has not been applied to extremely arid zones. The objectives of this study were to (1) evaluate the performance of RZWQM2 in simulating soil moisture content and crop production in an extremely arid area and (2) develop an optimal irrigation strategy using model-simulated crop water stress. In this study, RZWQM2 hybridized with DSSAT was calibrated and validated against soil moisture, cottield, and development stage data collected from 2006 to 2013 in a flood-irrigated cotton field located in an extremely dry oasis in Cele, situated in Xinjiang, China (mean annual precipitation 37 mm). The simulated water balance was analyzed to determine the actual crop water consumption, crop water requirements, and seepage loss. Subsequently, an optimal irrigation scheme was developed using RZWQM2 by averting crop water stress from planting to 90% open boll. In comparison to similar studies, the accuracy of soil moisture content simulations was deemed acceptable based on percent bias (PBIAS < ±15%), coefficient of determination (0.378 = R2 = 0.636), Nash-Sutcliffe model efficiency (0.130 = ME = 0.557), and root mean squared error (0.022 m3 m-3 = RMSE = 0.031 m3 m-3). The model performed well in simulating cotton yield (R2 = 0.79, ME = 0.75, RMSE = 417.0 kg ha-1, and relative RMSE (rRMSE) = 12.5%). Model-simulated plant emergence dates were generally six days late because of the model’s lack of a component for mulching after seeding. Other phenological dates were closely matched, with a mean difference of ±4 days. On average, over eight years, the simulated growing season (planting to 90% open boll) water balance showed that the cotton crop consumed 532 mm year-1 of water under current irrigation practices, while 109 mm of water was lost through deep seepage. However, based on simulated PET, the crop water requirement was 641 mm year-1, suggesting water stress under current irrigation practices. Under these conditions, water stress occurred mainly during the late stages of cotton growth. The model-simulated actual evapotranspiration (ET) is comparable to the calculated ET using the water balance method, with percent error of -1.3%, indicating the rationality of applying model-simulated results in a water stress-based irrigation scheduling method. On average, the water stress-minimizing RZWQM2 irrigation schedule resulted in an apparent irrigation water savings of 32 mm year-1 (4.9%) and an annual yield increase of 527 kg ha-1 (16.3%). RZWQM2 was shown to be suitable for simulating soil hydrology and crop development in an agricultural system implemented in an extremely dry climate. Rescheduling of irrigation using a water stress-based method can be used to optimize irrigation water use and cotton production.
Keywords.Cotton production, Optimum irrigation, RZWQM2, Soil water content, Water stress, WS-based regime.
Regions under a desert climate usually experience precipitation on the order of 25 to 200 mm year-1 (Laity, 2009) and are subject to significantly greater potential evaporation (Ep). Located in northwest China’s desert climate zone, the Xinjiang district produces almost two-thirds (62.5% and 67.3% in 2015 and 2016, respectively) of China’s total cotton production (China Agriculture, 2016) but at the cost of the agricultural sector accounting for 96.2% of the region’s total water usage (Karthe et al., 2015). Given agricultural productivity’s strong dependence on water availability, irrigation management is a critical practice in desert climate agroecosystems. Largely based on farmer’s confidence and experience with traditional irrigation methods, approximately 40% of fields in the Xinjiang district receive flood irrigation (Zhou et al., 2013), potentially resulting in substantial deep seepage losses and inefficient use of water resources. It is therefore critical to optimize water resource use by improving the cotton production process in terms of water use efficiency (WUE).
An accurate assessment of crop water requirements is an indispensable component in developing irrigation schemes that optimize water use. Crop water requirements can be measured using weighing lysimeters and water balance methods, but such methods are laborious and time-consuming. Liu et al. (2006) conducted a cotton irrigation experiment in Weili County, southern Xinjiang, and reported a cotton crop water requirement of 640 mm during the growing season, whereas Cai et al. (2002) found cotton water requirements to be about 380 mm in Shihez, northern Xinjiang. However, each of these measurements was based on a single cropping year. Another experiment on a cotton plantation was conducted in Cele (data used in this study) over a longer period (2006-2013); however, irrigation water requirement was not one of the factors investigated in that experiment, and the cotton crop could have suffered from water stress.
Agricultural system models are promising tools for simulating water balance and can be used to estimate potential evapotranspiration (ETp), in other words, long-term crop water requirements under no water stress conditions. The Root Zone Water Quality Model 2 (RZWQM2) is a processing model incorporating energy, water and nutrient equilibrium, plant growth, pesticide processes, and agricultural management components that has performed well in simulating WUE and crop productivity (Ahuja et al., 2000). Using this model, Qi et al. (2013) estimated growing season ETp for spring wheat (Triticum æstivum L.) grown in Sidney, Montana, to be 558 mm and suggested that an additional 323 mm of irrigation water should be applied to meet the crop’s water consumption in this region. Ma et al. (2012b) applied a parameterized RZWQM2 model to simulate maize responses to irrigation amounts representing specific percentages of estimated crop ETp and demonstrated the feasibility of using the model to schedule irrigation based on crop water requirements. Fang et al. (2010) evaluated RZWQM2 in simulating crop yield and soil water balance responses to different irrigation treatments and investigated irrigation strategies to reach high yield and WUE based on model simulation. This model has been well evaluated in terms of simulating water balance and crop production in many previous studies in many different field conditions. When used for simulating various variables (e.g., crop production, energy balance, water stress, and crop response to climate change under full and deficit irrigation) for conditions prevailing in Colorado (Qi et al., 2016), the RZWQM2 model showed itself capable of estimating the effects of water stress on yield, water and energy balances. The model was also used to develop new crop cultivars (Ma et al., 2017). In the U.S. Central Great Plains, Saseendran et al. (2013) employed RZWQM2 to simulate the long-term yield responses of several plants under various levels of soil available water at planting and then used the results to assess the potential to increase cropping frequency in a wheat-fallow rotation system. Their results showed RZWQM2 to be accurate in simulating soil moisture, crop production, and water balance; however, RZWQM2 has never been evaluated under an extremely arid climate. Accordingly, in the present study, the model was evaluated against data collected from an extremely arid oasis in Xinjiang.
Water stress as simulated using RZWQM2 can poten-tially be used to trigger irrigation, thereby informing the development of an optimal irrigation schedule. Fang et al. (2007) demonstrated that wheat grain yield and WUE were significantly affected by crop water stress. Bausch et al. (2011) employed the ratio of stressed to non-stressed canopy temperature to quantify crop water stress. However, this method requires the co-existence of both full and deficit irrigation treatments. A recent study showed that water stress and crop yield response to water stress can be accurately simulated using a well-calibrated RZWQM2 model (Qi et al., 2016) and that these could, in turn, serve in irrigation scheduling, resulting in water savings for a corn field in Colorado (Gu et al., personal communication, 2017). Because no well-defined irrigation scheme exists for cotton cultivation in Xinjiang, the present study was designed to (1) evaluate the applicability of RZWQM2 in simulating soil moisture, cotton yield, phenology, and water balance under extremely arid condition, and (2) develop an optimal irrigation strategy based on simulated crop water stress using the calibrated RZWQM2 model.
Materials and Methods
Located in a warm temperate continental desert climate district situated in the center of the Eurasian continent (36°59' N, 80°48' E), the Cele oasis constitutes an extremely arid environment, with sparse precipitation, adequate light, a large diurnal temperature range, but also a long frost-free period (210 days) conducive to the growth of various crops, and relatively abundant light and heat resources (Zeng, 1999). The long-term mean air temperature is 11.9°C, and the percentage of available sunshine days exceeds 60%. The main type of surface soil in the oasis farmland is sandy loam soil (classification standard defined by USDA) (Zeng et al., 2010). Cotton fields in the Cele district are irrigated with Cele River floodwaters or underground water (Li et al., 2011).
The long-term (1960-2007) mean precipitation at the Cele oasis meteorological station is 37 mm year-1, while pan evaporation is 2729 mm year-1. The high potential evaporation capacity and limited precipitation often lead to low soil moisture (?) conditions. Agriculture is the main economic driver in the Cele oasis ecosystem region, with agriculture accounting for 65.63% of the total economic output in this area. Cotton is one of the most important cash crops in the Cele oasis.
The rectangular experimental plot measures 150 m (north-south) ×140 m (east-west) and was planted with cotton. Planting density was 23 to 45 plants m-2 with a maximum plant height of approximately 75 cm and rooting depth of 100 cm. The experimental field was equipped with neutron probe (CNC100, Probe Science and Technology Ltd., Beijing, China; previously calibrated for the studied soil) access tubes to measure volumetric ?. Two access tubes were installed in this field along the east-west center line. Each tube was about 35 m from the edge and the center of field, representing one side of the field. Volumetric water content (?) was measured every 0.10 m to a depth of 0.60 m and then every 0.20 m to a depth of 1.00 m every five days from 1 January 2007 to 31 December 2013. Measured ? values from the two tubes were averaged to represent the soil water in the research field, where the water table depth exceeded 15 m, so groundwater had little if any impact on ? in the upper layers of the agricultural soil (Gui et al., 2009).
The timing of phenological stages of cotton and estimation of seed yield were conducted in this field during the growth period and after harvest, respectively, during every year from 2006 to 2013. The cultivars Ceke No. 1 and Xinlu No. 21 were grown in 2006-2011 and 2012-2013, respectively. Field management information, such as planting and harvest dates, fertilization, and irrigation data, were also recorded. Flood irrigation was used in this field, and ridges were made in this field for more uniform application of flood irrigation. The timing and amounts of irrigation are listed in table 1. Meteorological data, including air temperature, wind speed, radiation, pan evaporation, relative humidity, photosynthetically active radiation (PAR), and rainfall were collected hourly at a meteorological station situated 20 m away from the research site. To determine yield, cotton in four randomly selected 1.0 m × 1.0 m squares was sampled. Total cotton, separated fiber, and seed were dried at 60°C for three days prior to weighing. Only the seed yield was used to calibrate and validate the model.
Table 1. Irrigation amounts and frequency in the cotton field from 2006 to 2013. Year Irrigation Events Total 1 2 3 4 5 6 7 2006 Date 31 Mar. 01 June 14 June 30 June 17 July 02 Aug. 25 Aug. 7 events Amount (cm) 15 12 12 12 12 12 12 87 2007 Date 01 Apr. 08 June 27 June 15 July 05 Aug. 25 Aug. 6 events Amount (cm) 15 9 9 9 9 9 60 2008 Date 02 Apr. 14 June 27 June 28 July 07 Aug. 5 events Amount (cm) 15 9 9 9 9 51 2009 Date 03 Apr. 29 May 14 June 27 June 28 July 21 Aug. 6 events Amount (cm) 15 12 12 12 12 12 75 2010 Date 07 Apr. 14 June 03 July 15 July 13 Aug. 04 Sept. 6 events Amount (cm) 15 10 10 10 10 10 65 2011 Date 15 Apr. 08 June 26 June 18 July 17 Aug. 13 Sept. 6 events Amount (cm) 15 10 10 10 10 10 65 2012 Date 11 Apr. 03 June 26 June 12 July 03 Aug. 23 Aug. 6 events Amount (cm) 15 10 10 10 10 10 65 2013 Date 16 Apr. 19 June 02 July 20 July 16 Aug. 5 events Amount (cm) 15 10 10 12 12 59
RZWQM2, developed by USDA Agricultural Research Service (ARS), is a one-dimensional model that includes modules for hydrology, energy balance, water quality, and crop growth. It has gained wide popularity in developing and evaluating management practices (Ma et al., 2012a). The DSSAT crop models were incorporated into RZWQM2 to better simulate crop growth and development (Jones et al., 2003; Ma et al., 2006a, 2006b). More functions have been incorporated into RZWQM2 over the years, and the model can now be applied in simulating evapotranspiration, infiltration and soil water redistribution, subsurface drainage, organic matter and nutrient (N) cycling, as well as the fate and transport of pesticides. Ma et al. (2012b) verified the feasibility of applying RZWQM2 in simulating the response of maize growth to irrigation in Colorado; they concluded that the parameterized RZWQM2 was capable of simulating crop growth under different irrigation treatments, and the results could be used to schedule irrigation. Sassendran et al. (2010) used RZWQM2 to predict the rotation effect on crop production under semiarid conditions based on 14 years of observations of crop yield and biomass in a wheat-corn-millet rotation system. Qi et al (2013) quantified the influences of management practices on soil moisture and crop production based on the simulation results of RZWQM2 and extended the application to different meteorology and management conditions.
In order to simulate water transport in RZWQM2, water flow has been divided into two phases. The infiltration from rainfall or irrigation is modeled based on the Green-Ampt approach (Ahuja et al., 1995):
V = infiltration rate (cm h-1)
Ks = saturated hydraulic conductivity (cm h-1)
tc = capillary suction head (cm)
H0 = depth of ponding (cm)
Zwf = depth of wetting front (cm).
The water redistribution in the soil profile after infiltration is calculated with the Richards equation (Celia et al., 1990):
? = volumetric soil water content (cm3 cm-3)
z = soil depth (cm)
h = matric water head (cm)
K = hydraulic conductivity (cm h-1)
T = time (h).
RZWQM2 uses ET modeling to predict daily potential evaporation on a bare soil surface and canopy transpiration based on minimal surface resistances (Farahani and DeCoursey, 2000). The canopy and bare soil surface are divided into two layers in the ET model (Farahani and Ahuja, 1996), and the sum of latent heat for transpiration (?T) and evaporation (?Es) is regarded as potential evapotranspiration, which is estimated by the ET model using a revised form of the Penman-Monteith (P-M) ET equation (Shuttleworth and Wallace, 1985), written as:
?ET = ?T + ?Es (3)
In equation 3, ?T and ?Es can be expressed as:
Rn, G = net radiation and soil heat flux, respectively (W m-2)
? = rate of saturated vapor pressure change at certain temperature (kPa °C-1)
Rnsub = net radiation below the canopy (W m-2)
? = air density (kg m-3)
cp = specific heat of moist air (J kg-1 °C-1)
VPD = vapor pressure deficit measured at a reference level (kPa)
? = psychrometric constant (kPa °C-1)
rs, ra = resistance from canopy and bulk boundary layers, respectively (s m-1)
Rns, Gs = net radiation and soil heat flux on bare soil surface, respectively (W m-2).
Model Input, Calibration, and Validation
RZWQM2 was executed using hourly meteorological data from 2007 to 2013. Other input data included soil property data (bulk density and soil particle distribution) and field management data (crop planting, irrigation, and fertilization). The initial values for the soil hydraulic parameters were estimated based on soil texture and bulk density (Schaap et al., 1998), and the range for each parameter based on sandy loam is given by Rawls et al. (1982). Subsequently, those parameters were calibrated manually against observed ? in different layers. The ? data for the years 2009, 2010, and 2013 were used for calibration, and the data for the other years were used for validation. The soil property information together with calibrated soil hydraulic parameters are listed in table 2.
Table 2. Observed bulk density, soil texture, and calibrated soil hydraulic parameters.[a] Layer
Soil Texture Ksat
Soil Moisture (m3 m-3) at Different Matric Potentials Sand
?m = 0
?m = -10 kPa
?m = -33 kPa
?m = -1500 kPa
?m = -8
0-0.15 1.40 66.1 25.0 8.9 2.59 0.57 51.24 0.300 0.214 0.122 0.0387 0.028 0.15-0.30 1.45 65.4 27.7 6.9 2.59 0.47 46.53 0.353 0.257 0.161 0.0558 0.035 0.30-0.60 1.45 64.8 25.6 9.7 2.59 0.44 53.80 0.323 0.255 0.167 0.0643 0.041 0.60-0.90 1.48 67.6 24.5 7.8 2.59 0.58 67.72 0.403 0.331 0.188 0.0608 0.045 0.90-1.20 1.43 65.9 24.1 10.0 2.59 0.57 68.07 0.403 0.332 0.188 0.0577 0.041
[a] ? = bulk density; sand, silt, and clay fractions are based on USDA particle size ranges; Ksat = soil saturated hydraulic conductivity; ? = Brooks-Corey pore size distribution index; pb = bubbling pressure; ?sat = saturated soil moisture content; ?fc* = soil moisture at field capacity, sandy soil; ?fc = soil moisture at field capacity, standard soil; ?pwp = soil moisture content at permanent wilting point; and ?r = residual water content.
Crop growth and development parameters were manually calibrated against observed phenology dates and yield (2006 to 2013) following the protocol suggested by Ma et al. (2011). Data for 2009, 2010, and 2013 were chosen for calibration, and the other years were used for validation. Available plant development-related data included planting date, flower emergence date, flowering date, boll cracking date, and harvest date. The two cultivars planted (Ceke No. 1 and Xinlu No. 21) were simulated using the initial parameters of HO0001 ACALA SJ-2CDM and GA0002 Georgia King in the incorporated DSSAT crop models, respectively. The crop parameters were manually calibrated against observed phenological dates and yield. The calibrated crop parameters for the two cultivars are listed in table 3.
Table 3. Calibrated crop development parameters for two cotton cultivars. Parameter Description Calibrated Value by Cultivar Ceke No. 1 Xinlu No. 21 EM-FL Time between plant emergence and flower appearance (days) 36 35 FL-SH Time between first flower and first pod (days) 5 11 FL-SD Time between first flower and first seed (days) 10 17 SD-PM Time between first seed and physiological maturity (days) 30 25 FL-LF Time between first flower and end of leaf expansion (days) 47 51 LFMAX Maximum leaf photosynthesis rate at 30°C, 350 vpm CO2, and highlight (mg CO2 m-2 s-1) 1.2 1.3 SLAVR Specific leaf area of cultivar under standard growth conditions (cm2 g-1) 200 220 SIZLF Maximum size of full leaf (cm2) 240 250 XFRT Maximum fraction of daily growth that is partitioned to seed + shell 0.67 0.85 WTPSD Maximum weight per seed (g) 0.2 0.2 SFDUR Seed filling duration for pod cohort at standard growth conditions (days) 18 18 SDPDV Average seeds per pod under standard growing conditions (seeds pod-1) 15 20 PODUR Time required for cultivar to reach final pod load under optimal conditions (days) 8 8
Because there was no field-observed cotton ET with which to evaluate the model-simulated actual ET, the cotton ET was estimated based on the neutron probe measured soil water depletion over the effective root zone soil depth. The following equation was applied to calculate cotton ET (mm) (Jensen et al., 1990):
n = number of soil depth increments (mm)
?1 and ?2 = volumetric soil water content on two separate sampling dates at soil depth I (cm3 cm-3)
?Si = thickness of each depth increment (mm)
I = depth of irrigation applied between the two sampling dates (mm)
R = depth of precipitation occurring between the two sampling dates (mm).
The maximum effective root zone soil depth was assumed to be 1.8 m (Bucks et al., 1988), and we also assumed that there was no deep seepage beyond 1.8 m soil depth. Other details are described by Hunsaker et al. (1994).
In this study, percent bias (PBIAS), coefficient of determination (R2), Nash-Sutcliffe model efficiency (ME), and root mean squared error (RMSE) were applied to evaluate the performance of the model simulation for soil water content, while for cotton yield the relative error (rRMSE) was also used:
n = total number of items in the data set
and = average measured and simulated values, respectively
Oi = ith measured value
Si = ith simulated value.
Model performance is considered acceptable if -15% < PBIAS < 15%, R2 > 0.70, ME > 0.5, and rRMSE < 30% (Hanson et al., 1999; Ahuja et al., 2000; Moriasi et al., 2007). For soil water content, the above statistics were compared with values from other similar studies. We chose the years 2009, 2010, and 2013, with relatively high production, for calibration and used the other years for validation. For the phenology dates, only the differences between simulated and observed dates were compared.
Crop Water Requirement
The ETp simulated by the calibrated RZWQM2 model during the growing season (from planting to 90% open boll) was treated as the crop water requirement. The water balance components, including soil evaporation, plant transpiration, and deep seepage, were analyzed to investigate soil water loss under flood irrigation.
Water Stress Based Irrigation Scheduling
The optimum irrigation schedule was generated using the calibrated RZWQM2 model, and the simulated water stress index was used to trigger irrigation. That is, when the model predicted any water stress, irrigation was applied to relieve the water stress, and the application rate was computed by the model to replenish the soil to ?fc. The water stress index used was the turgor factor (TURFAC) which is generated by the DSSAT-CSM crop growth model embedded in RZWQM2. TURFAC describes the level of plant water stress and relates to the expansion of plant leaf cells. It is calculated as (Saseendran et al., 2014, 2015):
TpSW = potential transpiration computed using the Shuttleworth-Wallace method in the model (mm)
RWU = DSSAT potential root water uptake (mm)
RWUEP1 = species-specific parameter used for evaluating water stress impact on expansion growth of cells (set at 1.5 for cotton).
Water stress occurs when RWU (supply) is not sufficient for crop TpSW (demand) estimated by the Shuttleworth-Wallace model. The value of TURFAC ranges from 0.0 (zero) for a fully water-stressed condition, to 1.0 for a water stress-free condition. Once water stress has occurred, the quantity of water to apply was calculated as:
?fcj = volumetric soil water content at field capacity of the jth soil layer (m3 m-3)
?t0j = simulated volumetric soil water content of the jth soil layer on the day of irrigation (m3 m-3)
IRt0 = irrigation water supply required on the day of irrigation (t0) (mm)
Dj = delineated jth soil layer depth (mm)
N = number of layers in the root zone from the soil surface to the rooting depth on the day of irrigation.
According to simulations run for irrigation treatments observed in the field, the growth of cotton almost ends after the “90% open boll” growth stage is reached. Under optimal irrigation scheduling, water application ends after the simulated “90% open boll” occurs. Irrigation applied prior to the planting date (pre-irrigation) was kept unchanged in the optimization. The minimum irrigation depth for each event was set to 50 mm, a practical number for flood irrigation in this sandy soil area.
To facilitate the optimization process, a Java program was developed to run the model automatically and then extract the simulated results and derive irrigation events before “90% open boll” in DSSAT using the method described. The Java program then wrote the derived irrigation event into the RZWQM2 input file automatically and ran the model again. This process was repeated until no water stress occurred before “90% open boll”.
Table 4. Statistics for observed and simulated values of soil moisture and evaluation parameters.[a] Depth
Calibration Phase Validation Phase Obs. Sim. PBIAS R2 ME RMSE rRMSE Obs. Sim. PBIAS R2 ME RMSE rRMSE 0.10 0.059 0.058 0.016 0.378 0.314 0.031 0.525 0.065 0.059 0.092 0.500 0.448 0.022 0.338 0.20 0.073 0.082 -0.123 0.538 0.330 0.025 0.342 0.081 0.087 -0.078 0.537 0.306 0.026 0.321 0.30 0.084 0.088 -0.046 0.602 0.433 0.023 0.274 0.089 0.093 -0.044 0.605 0.439 0.022 0.247 0.40 0.096 0.095 0.011 0.594 0.470 0.023 0.240 0.099 0.102 -0.034 0.597 0.413 0.023 0.232 0.50 0.106 0.099 0.067 0.615 0.510 0.024 0.226 0.108 0.106 0.016 0.513 0.310 0.025 0.231 0.60 0.111 0.100 0.093 0.594 0.453 0.024 0.216 0.116 0.107 0.072 0.636 0.412 0.022 0.190 0.80 0.126 0.121 0.042 0.451 0.349 0.033 0.262 0.126 0.121 0.016 0.612 0.340 0.023 0.183 1.00 0.121 0.127 -0.051 0.592 0.557 0.026 0.215 0.130 0.135 -0.039 0.594 0.130 0.024 0.185
[a] Obs. = observed mean (m3 m-3), Sim. = simulated mean (m3 m-3), PBIAS = percent bias, R2 = coefficient of determination, and ME = Nash-Sutcliffe model efficiency.
Irrigation water use efficiency (IWUE), representing the effectiveness of irrigated water in generating yield, was calculated as (O’Shaughnessy et al., 2012):
Iw = amount of irrigation water (mm)
IWUE = irrigation water use efficiency (kg ha-1 mm-1)
Y and Yo = yield under irrigation management and under rainfed production, respectively (kg ha-1).
The water savings ratio, yield increase, and IWUE increase under irrigation management (vs. rainfed production) were calculated as follows:
?I, ?Y, and ?IWUE = irrigation water savings ratio (%), yield increase (%), and IWUE increase (%), respectively
IFO, YFO, and IWUEFO = irrigation water quantity (mm), yield (kg ha-1), and IWUE (kg ha-1 mm-1), respectively, measured under current irrigation management practices
IWS, YWS, and IWUEWS = irrigation water quantity (mm), yield (kg ha-1), and IWUE (kg ha-1 mm-1), respectively, simulated as occurring under the water stress-based irrigation method.
Results and Discussion
The statistical parameters used to evaluate model performance in simulating ? are presented in table 4. For all soil layers, the simulated ? showed an acceptable level of agreement with observed data in both calibration and validation (PBIAS < ±15%, 0.378 = R2 = 0.615, 0.314 = ME = 0.557, and 0.023 = RMSE = 0.033 for calibration, and PBIAS < ±10%, 0.500 = R2 = 0.636, 0.130 = ME = 0.448, and 0.022 = RMSE = 0.026 for validation). Although these statistics did not meet the “satisfactory” criterion, particularly for the top layers, given the difficulty in simulating soil water content, our results are comparable to those obtained from research conducted on the North China Plain (Fang et al., 2014a), where the statistics were -0.44 = ME = 0.25 and 0.30 = R2 = 0.66 across different layers, and the results from Fang et al. (2014b), where ME varied from -2.263 to 0.203. The rRMSE for different soil layers was mostly “satisfactory” and varied from 0.183 to 0.525.
Because there was no recharge from groundwater, precipitation and irrigation were the main sources of moisture in this district. For the upper soil layers, simulated ? values tended to have sharper peaks (fig. 1), which may because of lower ?fc (table 2) and therefore less water storage. The ? showed a roughly increasing trend within each soil increment from the surface to 1.00 m depth, indicating better water retention capacity in the deeper soil profile. Such results have also been observed and explained by Li et al. (2011), whose study was conducted on neighboring farmland.
With the exception of the emergence dates, simulated dates for phenological stages (table 5) were, in general, within four days of the observed dates. The simulated results for the “flowering” and “bolls cracked” dates were satisfactory, with average deviations of -3.9 days and 0.9 day, respectively. Simulated emergence dates were on average six days later than the observed dates. This can be explained by the fact that the plastic mulch accelerated germination through heat conservation, whereas RZWQM2 is not yet able to simulated mulching. Notably, in 2006 and 2011, the simulated flowering dates were significantly earlier than the observed dates, which may be the result of an inaccurate assessment of phenological stages in those years.
The cotton seed yield was, in general, adequately simulated (PBIAS = -0.0249, R2 = 0.79, and ME = 0.75) from 2006 to 2013. The RMSE and rRMSE values for yield simulation were 417.0 kg ha-1 and 12.5%, respectively, similar to those reported by Anapalli et al. (2016), where the RMSE and rRMSE were 333 kg ha-1 and 14%, respectively. Specifically, the results for the calibration years (PBIAS = -0.0869, R2 = 0.93, ME = 0.71, RMSE = 499.7 kg ha-1, and rRMSE = 12.3%) were comparable to those for the validation years (PBIAS = 0.0271, R2 = 0.75; ME = 0.78, RMSE = 358.3 kg ha-1, and rRMSE = 12.4%). The significant overestimation (+30.1%) of yield in 2008 (table 6) might be attributable to the milder weather conditions and relatively lower ETp in 2008 (table 7). Moreover, the total depth of irrigation in 2008 was 36 mm, which is much lower than in other years. Although no observations of water stress were made, presumably a crop growing under a relatively high water stress would
Figure 1. Simulated and observed soil moisture content within 0 to 100 cm soil depth from 2007 to 2013.
show a significant yield loss. The fact that the simulated yield reduction in 2008 was not as great as that observed may suggest that the modeled response of cotton yield to high water stress needs further investigation. In general, the results indicated that RZWQM2 was capable of simulating crop phenology and yield under the extremely arid climate at the research site.
Based on the statistics in figure 2, the model slightly underestimated evapotranspiration during the cotton growing season, with a percent error of -1.3%. Relatively larger dif-erences can be noticed in the years with higher production: 2010 (-9.8%) and 2013 (-10.8%). These evaluation statistics are comparable to the research by Thorp et al. (2014), in which absolute error varied from -42 to +63 mm and percent error varied from -6% to +7% in different experiments, indi-cating that the simulated results can be applied to estimate cotton water requirements and optimize irrigation practices.
Table 6. Simulated and observed cotton seed yield and yield difference from 2006 to 2013. Cultivar Year[a] Yield
Obs. Sim. Ceke No. 1 2006 (V) 2857 3179 11.3 2007 (V) 2716 2427 -10.6 2008 (V) 1932 2513 30.1 2009 (C) 3226 3323 3.0 2010 (C) 4344 3963 -8.8 2011 (V) 3067 2741 -10.6 Xinlu No. 21 2012 (V) 3880 3984 2.7 2013 (C) 4570 3799 -16.9
[a] C = calibration; V = validation.
Table 7. Parameters in water balance equation during crop growing season (from planting to 90% bolls open) for 2006 to 2013. Year Water Balance Parameters (mm)[a] ?S P I Ds Ea Ta ETa Ep Tp ETp D 2006 -105.12 14.0 720 144.0 145 555 699 449 581 1030 331 2007 -64.97 14.0 450 19.5 63.8 447 511 173 521 694 183 2008 -117.64 34.6 360 22.4 60.7 430 491 123 493 617 126 2009 -45.46 37.4 600 138.0 74.9 471 546 111 528 639 93 2010 -88.57 77.6 500 147.0 74.1 446 520 88 457 545 25 2011 -134.55 3.0 400 49.4 56.6 432 489 85 473 558 69 2012 -78.18 58.0 500 130.0 60.2 446 506 69 446 515 9 2013 -127.54 28.6 440 102.0 53.1 442 495 68 466 534 39 Mean -91.34 33.4 496 94.1 73.5 459 532 146 496 641 109
[a] ?S = change of soil water storage in the soil profile; P = cumulative precipitation; I = cumulative irrigation; Ds = cumulative deep seepage out of the soil profile; Ea = cumulative actual evaporation; Ta = cumulative actual transpiration; ETa = cumulative actual evapotranspiration; Ep = cumulative potential evaporation; Tp = cumulative potential evaporation; ETp = cumulative potential evapotranspiration; and D = difference between ETp and ETa. The I in this table did not include pre-irrigation before planting (15 cm for each year) and irrigation after 90% open boll in 2011 (10 cm).
Simulated Crop Water Requirement
The model-simulated water balance components (table 7), excluding surface runoff, which was zero for all the years due to the oasis soil’s sandy texture and high Ksat (table 2), show irrigation to be the main water input, a large portion of which is lost to ET. Crop water requirements, which are equivalent to ETp, were 641 mm during the cotton growing season. Although the field-observed ET was not recorded to perform the validation, this simulated cotton water requirement is comparable to the results from research conducted in south Xinjiang, where the water demand of cotton was 625 mm (Liu et al., 2006). Under crop management conditions based on farmer experience, about 496 mm of water was applied. Taking the mean growing season rainfall of 33.5 mm into account, an additional 109 mm water should be applied to meet the crop water requirement. According to the simulation, under current flood irrigation practices, a substantial quantity of water (94.1 mm year-1 on average, or 17.8% of the overall water supply) is wasted through deep seepage.
The simulation showed a reasonable cotton water requirement compared to other studies. From the analysis of water supply, the simulated ETa for the crop growing period represented from 67.9% to 98.3% of ETp, with an average of 79% (ETa = 508 mm). In the present study, there were more substantial differences between ETp and ETa in the first three years (2006-2008), when water stress was suspected. The relatively lower observed yield and the simulated lower ETa to ETp ratio confirmed this suspicion. In contrast, the years with higher production had relatively lesser differences between ETp and ETa, suggesting that the applied water may have closely met the crop water requirement.
The simulation further suggests that, although rainfall was consistently low over all years, variation in other climate factors (e.g., solar radiation and relative humidity) may result in significantly different crop water requirements in an extremely arid climate. In 2006, for example, the irrigation depth was greatest among all years, whereas the yield was not correspondingly high, indicating that the quantity of irrigation alone did not meet crop water requirement, which was also greatest among all years. The simulation also justifies the idea that if a long-term field experiment is not available, a long-term simulation is likely to be more reliable than a one-year field study for determining crop irrigation requirements.
Table 8. Comparison of the irrigation and yield response between current irrigation management and water stress-based irrigation. Year Current Irrigation Management RZWQM2-Derived Water Stress-Based Irrigation[a] Irrigation
(kg ha-1 mm-1)
Irrigation / ?I
(mm / %)
Yield / ?Y
(kg ha-1 / %)
IWUE / ?IWUE
(kg ha-1 mm-1 / %)
2006 870 3179 3.65 744 / 14.5 3184 / 0.2 4.28 / 17.1 2007 600 2427 4.01 673 / -12.1 3310 / 36.4 4.92 / 21.6 2008 510 2513 4.93 678 / -32.9 3787 / 50.7 5.59 / 13.4 2009 750 3323 4.43 668 / 11.0 4489 / 35.1 6.72 / 51.7 2010 650 3963 6.1 532 / 18.1 3878 / -0.2 7.29 / 19.6 2011 650 2741 4.22 599 / 7.8 3657 / 33.4 6.11 / 44.8 2012 650 3984 6.13 521 / 19.9 3983 / 0.0 7.64 / 24.7 2013 590 3799 6.44 599 / -1.5 3855 / 1.5 6.44 / 0.0 Mean 659 3241 4.92 627 / 4.9 3768 / 16.3 6.01 / 22.2
[a] Based on TURFAC (eq. 12); IWUE = irrigation water use efficiency (eq. 14), ?I = percent irrigation water savings (eq. 15), ?Y = percent yield increase (eq. 16), and ?IWUE = percent IWUE increase (eq. 17).
Optimum Irrigation Scheduling
A water stress-based irrigation scheduling method based on RZWQM2-simulated water stress was applied to optimize the irrigation schedule. The original field records and optimized data are presented in table 8 and figure 3. In general, the irrigation optimization approach suggests that applying 616 mm water to the cotton crop over the growing season allowed the crop to achieve its highest yield potential (3.777 Mg ha-1). The optimized irrigation treatment resulted in almost the same yield as the current practice in 2006, 2010, 2012, and 2013 and provided 26.5% and 42.3% greater yields in 2011 and 2008, respectively. The quantity of model-suggested irrigation water applied was lower than the amount applied under current practices in 2012 and 2013 (-22.9% and -2.6%, respectively) but higher in 2007 (+14.9%) and 2008 (+32.5%). Across all eight years, optimized irrigation resulted in an increase in IWUE ranging from 2.4% (2013) to 50.9% (2009), with a mean increase of 22.2%. Even for years with no yield increase or water savings under optimized irrigation, IWUE was improved (table 8), indicating that the current irrigation management either applied insufficient water, resulting in yield loss, or applied more water than the crop required to obtain a high yield. The water stress-based method proved to be an efficient way to schedule (based on a calibrated RZWQM2 model) high-yield and water-saving irrigation.
With 59 irrigation events under the water stress-based method (fig. 3), compared to 47 events under the current practice, water stress prior to the “90% open boll” stage was eliminated in all eight years. Although the irrigation frequency was higher with the water stress-based method (i.e., an average of eight extra irrigation events per year), the water applied during each event was decreased to only the amount (>50 mm) required to replenish the soil in the root zone to ?fc. This amount ensured the crop water requirement while minimizing deep seepage. The water quantity at each event under the current irrigation practice was apparently greater than the capacity of the soil in the root zone; therefore, excessive irrigation water was wasted through deep seepage.
For the years 2006, 2010, 2012 and 2013, yields remained largely unchanged when shifting from the current practice to water stress-based irrigation scheduling (table 8), In those years, crop water stress under current irrigation management occurred just before the “90% open boll” stage, so any yield change might have been caused by the interactive effects of water and nitrogen stresses. Irrigation is required in extremely arid areas like Xinjiang, with a limited water supply, to avoid the occurrence of crop water stress. To derive an optimal irrigation management strategy that both minimizes water consumption and allows an acceptable yield, growers should be able to precisely estimate current ? and crop growth conditions, including whether the crop is under water stress or nitrogen stress, what the current growth stage is, etc. By applying the calibrated RZWQM2 model, crop water stress, soil moisture conditions, and crop growth can be estimated accurately, and this information can be applied to derive an optimum irrigation schedule for this area in the future.
In the analysis of potential evapotranspiration, 109 mm more irrigation was recommended to meet 100% of the crop water requirement, while the water stress-based irrigation scheduling method saved irrigation by decreasing the water lost through deep seepage from 94.1 to 26.2 mm (table 9). The other parameters in the water balance equation did not vary much before and after applying the water stress-based scheduling method. On average, the water stress-minimizing RZWQM2 irrigation schedule resulted in an apparent water savings of 32 mm year-1 and an annual yield increase of 527 kg ha-1. Irrigation scheduling using a water stress-based method can be used to optimize water use and cotton production.
To obtain an optimum irrigation schedule for cotton under an extremely arid desert oasis climate, the performance of RZWQM2 was evaluated in simulating soil water content, crop production, and crop phenological stage against field-observed data from the Cele oasis-desert transition zone in Xinjiang, China. The cotton yield and phenology dates were in general satisfactorily simulated by the RZWQM2 model. The model performance under this climate can be deemed acceptable when compared to studies conducted in other regions. The simulated water balance showed that the cotton crop water requirement was 641 mm, and an additional application of 109 mm of irrigation, compared to the current farmer-directed irrigation of 500 mm, was needed to meet the crop water requirement under an average rainfall of 33.4 mm year-1. The simulation also suggested that current irrigation practices may result in a 94.1 mm loss of water through deep percolation. The irrigation schedule was optimized using model-simulated cotton crop water stress. The optimum irrigation schedule thereby derived could increase crop yield and irrigation water use efficiency by 16.3% and 22.2%, respectively, while reducing water application by 4.9%, compared to current irrigation management. This study demonstrated the model to be capable of simulating irrigated cotton production under an extremely arid desert oasis climate, and the model-simulated water stress could be used as a sufficient parameter for optimizing irrigation water use. It is further suggested that RZWQM2 could prove a useful tool for regional water planning, given its accurate simulation of crop water requirements.
Figure 3. Water stress (TURFAC) response under (a) current irrigation management and (b) water stress-based irrigation.
Table 9. Comparison of water balance during 2006 to 2013 growing seasons (planting to 90% open boll stage ) between current irrigation management and water stress (WS) based optimal irrigation. Irrigation
Water Balance Parameters (mm)[a] ?S P I Ds Ea Ta ETa Current -91.34 33.4 496 94.1 74 459 532 WS-based -87.88 33.4 477 26.2 65 499 564
[a] ?S = change of soil water storage in soil profile; P = cumulative precipitation; I = cumulative irrigation; Ds = cumulative deep seepage out of soil profile; Ea = cumulative actual evaporation; Ta = cumulative actual transpiration; and ETa = cumulative actual evapotranspiration.
Ahuja, L. R., Johnsen, K. E., & Heathman, G. C. (1995). Macropore transport of a surface-applied bromide tracer: Model evaluation and refinement. SSSA J., 59(5), 1234-1241. https://doi.org/10.2136/sssaj1995.03615995005900050004x
Ahuja, L. R., Rojas, K. W., Hanson, J. D., Shaffer, M. J., & Ma, L. (2000). Root zone water quality model. Highlands Ranch, CO: Water Resources Publications.
Anapalli, S., Pettigrew, W., Reddy, K., Ma, L., Fisher, D., & Sui, R. (2016). Climate-optimized planting windows for cotton in the lower Mississippi Delta region. Agronomy, 6(4), 46. https://doi.org/10.3390/agronomy6040046
Bausch, W., Trout, T., & Buchleiter, G. (2011). Evapotranspiration adjustments for deficit-irrigated corn using canopy temperature: A concept. Irrig. Drain., 60(5), 682-693. https://doi.org/10.1002/ird.601
Bucks, D. A., Allen, S. G., Roth, R. L., & Gardner, B. R. (1988). Short staple cotton under micro and level-basin irrigation methods. Irrig. Sci., 9(3), 161-176. https://doi.org/10.1007/bf00275429
Cai, H. J., Shao, G. C., & Zhang, Z. H. (2002). Water demand and irrigation scheduling of drip irrigation for cotton under plastic mulch. J. Hydraul. Eng., 33(11), 119-123.
Celia, M. A., Bouloutas, E. T., & Zarba, R. L. (1990). A general mass-conservative numerical solution for the unsaturated flow equation. Water Resour. Res., 26(7), 1483-1496. https://doi.org/10.1029/WR026i007p01483
China Agriculture. (2016). China’s cotton output continued its decline in 2016. Retrieved from http://www.chinaag.org/2016/12/30/chinas-cotton-output-continued-its-decline-in-2016/
Dourte, D., Bartel, R. L., George, S., Marois, J. J., & Wright, D. L. (2016). A sod-based cropping system for irrigation reductions. Renew. Agric. Food Syst., 31(6), 485-494. https://doi.org/10.1017/S1742170515000393
Fang, Q. X., Ma, L., Flerchinger, G. N., Qi, Z., Ahuja, L. R., Xing, H. T., ... Yu, Q. (2014b). Modeling evapotranspiration and energy balance in a wheat-maize cropping system using the revised RZ-SHAW model. Agric. Forest Meteorol., 194, 218-229. https://doi.org/10.1016/j.agrformet.2014.04.009
Fang, Q. X., Ma, L., Nielsen, D. C., Trout, T. J., & Ahuja, L. R. (2014a). Quantifying corn yield and water use efficiency under growth stage-based deficit irrigation conditions. In L. R. Ahuja, L. Ma, & R. J. Lascano (Eds.), Practical applications of agricultural system models to optimize the use of limited water (pp. 1-24). Madison, WI: ASA, CSSA, SSSA. https://doi.org/10.2134/advagricsystmodel5.c1
Fang, Q., Chen, Y., Yu, Q., Ouyang, Z., Li, Q., & Yu, S. (2007). Much improved irrigation use efficiency in an intensive wheat-maize double cropping system in the North China Plain. J. Integ. Plant Biol., 49(10), 1517-1526. https://doi.org/10.1111/j.1672-9072.2007.00559.x
Fang, Q., Ma, L., Yu, Q., Ahuja, L. R., Malone, R. W., & Hoogenboom, G. (2010). Irrigation strategies to improve the water use efficiency of wheat-maize double cropping systems in North China Plain. Agric. Water Mgmt., 97(8), 1165-1174. https://doi.org/10.1016/j.agwat.2009.02.012
Farahani, H. J., & Ahuja, L. R. (1996). Evapotranspiration modeling of partial canopy/residue-covered fields. Trans. ASAE, 39(6), 2051-2064. https://doi.org/10.13031/2013.27708
Farahani, H. J., & DeCoursey, D. G. (2000). Potential evaporation and transpiration processes in the soil-residue-canopy system. In L. R. Ahuja, K. W. Rojas, J. D. Hanson, M. J. Shaffer, & L. Ma (Eds.), Root zone water quality model (pp. 51-80). Highlands Ranch, CO: Water Resources Publications.
Gui, D., Lei, J., Mu, G., & Zeng, F. (2009). Effects of different management intensities on soil quality of farmland during oasis development in southern Tarim Basin, Xinjiang, China. Intl. J. Sustain. Devel. World Ecol., 16(4), 295-301. https://doi.org/10.1080/13504500903108887
Hanson, J. D., Rojas, K. W., & Shaffer, M. J. (1999). Calibrating the Root Sone Water Quality Model. Agron. J., 91(2), 171-177. https://doi.org/10.2134/agronj1999.00021962009100020002x
Hunsaker, D. J., Hendrey, G. R., Kimball, B. A., Lewin, K. F., Mauney, J. R., & Nagy, J. (1994). Cotton evapotranspiration under field conditions with CO2 enrichment and variable soil moisture regimes. Agric. Forest Meteorol., 70(1), 247-258. https://doi.org/10.1016/0168-1923(94)90061-2
Jensen, M. E., Burman, R. D., & Allen, R. G. (1990). Evapotranspiration and irrigation water requirements. Reston, VA: ASCE.
Jones, J. W., Hoogenboom, G., Porter, C. H., Boote, K. J., Batchelor, W. D., Hunt, L. A., ... Ritchie, J. T. (2003). The DSSAT cropping system model. European J. Agron., 18(3), 235-265. https://doi.org/10.1016/S1161-0301(02)00107-7
Karthe, D., Chalov, S., & Borchardt, D. (2015). Water resources and their management in central Asia in the early twenty-first century: Status, challenges, and future prospects. Environ. Earth Sci., 73(2), 487-499. https://doi.org/10.1007/s12665-014-3789-1
Laity, J. J. (2009). Deserts and desert environments (Vol. 3). Hoboken, NJ: John Wiley & Sons.
Li, H., Zeng, F., Gui, D., & Lei, J. (2011). Characteristics of soil environment variation in oasis-desert ecotone in the process of oasis growth. In D. Li, Y. Liu, & Y. Chen (Eds.), Computer and Computing Technologies in Agriculture IV (pp. 321-334). Berlin, Germany: Springer. https://doi.org/10.1007/978-3-642-18333-1_38
Liu, X. Y., Tian, C. Y., Ma, Y. J., & Liu, H. P. (2006). Water consumption characteristics and scheduling of drip irrigation under plastic film for cotton in south Xinjiang. Agric. Res. Arid Areas, 24(1), 108-112.
Ma, L., Ahuja, L. R., Islam, A., Trout, T. J., Saseendran, S. A., & Malone, R. W. (2017). Modeling yield and biomass responses of maize cultivars to climate change under full and deficit irrigation. Agric. Water Mgmt., 180(part A), 88-98. https://doi.org/10.1016/j.agwat.2016.11.007
Ma, L., Ahuja, L. R., Nolan, B. T., Malone, R. W., Trout, T. J., & Qi, Z. (2012a). Root Zone Water Quality Model (RZWQM2): Model use, calibration, and validation. Trans. ASABE, 55(4), 1425-1446. https://doi.org/10.13031/2013.42252
Ma, L., Ahuja, L. R., Saseendran, S. A., Malone, R. W., Green, T. R., Nolan, B. T., ... Hoogenboom, G. (2011). A protocol for parameterization and calibration of RZWQM2 in field research. In L. R. Ahuja & L. Ma (Eds.), Methods of introducing system models into agricultural research (pp. 1-64). Madison, WI: ASA, CSSA, SSSA. https://doi.org/10.2134/advagricsystmodel2.c1
Ma, L., Hoogenboom, G., Ahuja, L. R., Ascough, J. C., & Saseendran, S. A. (2006a). Evaluation of the RZWQM-CERES-Maize hybrid model for maize production. Agric. Syst., 87(3), 274-295. https://doi.org/10.1016/j.agsy.2005.02.001
Ma, L., Hoogenboom, G., Ahuja, L. R., Ascough, J. C., & Saseendran, S. A. (2006b). Evaluation of the RZWQM-CERES-Maize hybrid model for maize production. Agric. Syst., 87(3), 274-295. https://doi.org/10.1016/j.agsy.2005.02.001
Ma, L., Trout, T. J., Ahuja, L. R., Bausch, W. C., Saseendran, S. A., Malone, R. W., & Nielsen, D. C. (2012b). Calibrating RZWQM2 model for maize responses to deficit irrigation. Agric. Water Mgmt., 103, 140-149. https://doi.org/10.1016/j.agwat.2011.11.005
Moriasi, D. N., Arnold, J. G., Van Liew, M. W., Bingner, R. L., Harmel, R. D., & Veith, T. L. (2007). Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Trans. ASABE, 50(3), 885-900. https://doi.org/10.13031/2013.23153
O’Shaughnessy, S. A., Evett, S. R., Colaizzi, P. D., & Howell, T. A. (2012). Grain sorghum response to irrigation scheduling with the time-temperature threshold method and deficit irrigation levels. Trans. ASABE, 55(2), 451-461. https://doi.org/10.13031/2013.41395
Qi, Z., Bartling, P. N., Jabro, J. D., Lenssen, A. W., Iversen, W. M., Ahuja, L. R., ... Evans, R. G. (2013). Simulating dryland water availability and spring wheat production in the Northern Great Plains. Agron. J., 105(1), 37-50. https://doi.org/10.2134/agronj2012.0203
Qi, Z., Ma, L., Bausch, W. C., Trout, T. J., Ahuja, L. R., Flerchinger, G. N., & Fang, Q. (2016). Simulating maize production, water and surface energy balance, canopy temperature, and water stress under full and deficit irrigation. Trans. ASABE, 59(2), 623-633. https://doi.org/10.13031/trans.59.11067
Rawls, W. J., Brakensiek, D. L., & Saxtonn, K. E. (1982). Estimation of soil water properties. Trans. ASAE, 25(5), 1316-1320. https://doi.org/10.13031/2013.33720
Saseendran, S. A., Ahuja, L. R., Ma, L., Nielsen, D. C., Trout, T. J., Andales, A. A., ... Ham, J. (2014). Enhancing the water stress factors for simulation of corn in RZWQM2. Agron. J., 106(1), 81-94. https://doi.org/10.2134/agronj2013.0300
Saseendran, S. A., Nielsen, D. C., Ahuja, L. R., Ma, L., & Lyon, D. J. (2013). Simulated yield and profitability of five potential crops for intensifying the dryland wheat-fallow production system. Agric. Water Mgmt., 116, 175-192. https://doi.org/10.1016/j.agwat.2012.07.009
Saseendran, S. A., Nielsen, D. C., Ma, L., Ahuja, L. R., & Vigil, M. F. (2010). RZWQM2 simulations of alternative cropping systems with and without summer crops in the Central Great Plains. Agron. J., 102(5), 1521-1534. https://doi.org/10.2134/agronj2010.0141
Saseendran, S. A., Trout, T. J., Ahuja, L. R., Ma, L., McMaster, G. S., Nielsen, D. C., ... Ham, J. (2015). Quantifying crop water stress factors from soil water measurements in a limited irrigation experiment. Agric. Syst., 137, 191-205. https://doi.org/10.1016/j.agsy.2014.11.005
Schaap, M. G., Leij, F. J., & van Genuchten, M. Th. (1998). Neural network analysis for hierarchical prediction of soil water retention and saturated hydraulic conductivity. SSSA J., 62(4), 847-855.
Shuttleworth, W. J., & Wallace, J. S. (1985). Evaporation from sparse crops: An energy combination theory. Qtly. J. Royal Meteorol. Soc., 111(469), 839-855. https://doi.org/10.1002/qj.49711146910
Thorp, K. R., Barnes, E. M., Hunsaker, D. J., Kimball, B. A., White, J. W., Nazareth, V. J., & Hoogenboom, G. (2014). Evaluation of CSM-CROPGRO-Cotton for simulating effects of management and climate change on cotton growth and evapotranspiration in an arid environment. Trans. ASABE, 57(6), 1627-1642. https://doi.org/10.13031/trans.57.10612
Zeng, F. (1999). The countermeasures of suntainable development in Cele oasis ecosystem, Xinjiang. Arid Zone Res., 16(1), 29-34.
Zeng, F., Lei, J., & Zhang, X. (2010). Cele oasis: Environmental characteristics of desert transitional area and dominant plants adaptation [in Chinese]. Beijing, China: Science Press.
Zhou, H., Zhang, M., Zhou, Q., Sun, Z., & Chen, J. (2013). Analysis of agricultural irrigation water-using coefficient in Xinjiang arid region. Trans. CSAE, 29(22), 100-107.