Optimizing Rice Irrigation Strategies to Maximize Water Productivity: A Simulation Study Using AquaCrop Model for the Yanyun Irrigation District, Yangzhou, China

Abstract

The AquaCrop model is used to predict rice yield in response to different irrigation management in the Yanyun irrigation area in Yangzhou, China, and the constraints to rice production were identified to maximize water productivity based on model simulations. The model was calibrated by comparing measured and predicted canopy cover (CC), yield, and soil water content during the growing season in 2018. The results showed that, for CC simulations, R² was 0.99, RMSE was 3.6%, and NRMSE was 5.3%; for Biomass simulation, RMSE was 0.50 t/ha, and NRMSE was 5.3%. Different irrigation strategies were analyzed for a long-term simulation period from 1955 to 2014. The simulated rice yield increased rapidly as irrigation demand increased initially, and then gradually stabilized. The simulated rice yield fluctuated in the different years. The Pearson type-III model method was used to identify different hydrological years of wet, normal, and dry years. The analysis identified the wet year as 1991, normal year as 1981, and dry year as 1966. In the different rainfall years (1991, 1981, and 1966) water use efficiency (WUE), water productivity (WPet), and irrigation water productivity (IWP) were utilized to determine the irrigation strategy. The predicted highest WPet in the wet year was 1.77kg m⁻³, while the lowest WPet in the dry year was 1.13 kg m⁻³. The highest IWP was 19.78 kg m⁻³ in the wet year, and 9.32 kg m⁻³ in the normal year; while the lowest IWP in the dry year was 1.90 kg m⁻³. IWP was significantly higher in the rainy year, while WUE was significantly lower. On the other hand, WPet was more extensive in the wet year because the yield was higher, and the Evapotranspiration (ET) was smaller in comparison to the dry year.

1. Introduction

Rice (Oryza sativa L.) is one of the most widely grown crops globally, and it plays an essential role in global food security as the world population grows [1]. In China, rice is the second-largest crop production, and it accounts for 18.8% of the world’s rice area and 28.1% of global rice yields [2]. Rice is a wetland crop; it requires a ponding depth of water during most of the growing season and consumes a large amount of water due to evapotranspiration and deep percolation. In the major rice-growing area of southeastern China, rice irrigation requires a significant amount of water despite the humid climate; improving irrigation water use efficiency for rice production has a growing importance due to economic development and the rising concern of climate change [2].

Water requirements for rice are affected by multiple factors, including the cultivar, soil texture, field management, and meteorological factors during the rice-growing period [3]. The effect of variable weather conditions on rice irrigation water use can hardly be evaluated properly through field experiments. Properly calibrated crop models are often used for such purposes. These models can be applied to evaluate experimental studies and serve as agronomic research tools to synthesize research knowledge and decision support tools for system management [3].

The FAO AquaCrop model is relatively user-friendly in this context. It has been parameterized and used to simulate variable crop yield response to water worldwide [4]. Raes et al. [5] assessed the AquaCrop model for various crops. They found that the user can follow variations in soil water content changes related to crop development, evaporation, perspiration of soil, generation of biomass, and production development of yield. The model may be paused at any time to analyze the effect of changes in water-related inputs, making it suitable for designing deficit irrigation plans and scenario analysis. Maniruzzaman et al. [4] investigated AquaCrop’s performance under several water regimes in Bangladesh. They concluded that flooded paddies under continuous percolation performed well, and that AquaCrop could simulate seasonal field water balances. In a semiarid setting in Iran, Saadati et al. [3] used experimental data for rice crops to test the AquaCrop model under different irrigation management strategies. They found that under various irrigation managements strategies, the AquaCrop model can accurately predict rice’s canopy cover development and grain yield. The model can be used to investigate different management strategies for increasing rice water productivity [3]. At the same time, AquaCrop simulated the trend of the field water balance; the model accurately predicted rice CC, ET, biomass, and yield; these values were significantly underestimated around midseason. However, even during the recovery period, flooding water depth and soil moisture content were underestimated during rice season [2]. The AquaCrop model can simulate crop development over a long period using historical climate data. For Bolivia’s quinoa, practicable diagrams were established using long-range climate data series and analysis methods. Geerts et al. found that AquaCrop is a user-friendly and robust model that helps break the gap between farmers’ agricultural models that required sustainable irrigation guidelines [6]. García-Vila and Fereres revealed that the developed economic optimization method was a helpful tool to study simulation analyses and assist irrigation strategy managers, water authorities, and policymakers in making farm irrigation more sustainable [7].

Because the AquaCrop model can be a valuable tool for stakeholders involved in irrigated rice water management, it can provide insights into the performance of different water management strategies and help optimize crop yield and water use efficiency. In this paper, we test the applicability of AquaCrop for rice production in the Yanyun Irrigation area in Yangzhou, China and predict the water balance of paddy fields under different irrigation strategies, in addition to identifying the constraints on rice production to maximize water productivity.

2. Materials and Methods

2.1. Site Description and Data Collection

The simulation study was based on weather and crop data collected from an experimental station (longitude 119°30′ E, latitude 32°33′ N, and altitude 5.4 m) located in the Yanyun irrigation area on the lower reaches of the Yangtze River basin in Jiangsu, China. The annual average precipitation is 1020.1 mm, the average temperature is 15.6 °C, and the frost-free period is 221.7 d in the study area. A weather station (CR1000, Campbell Scientific, Jiangsu, China) has collected daily rainfall, air temperature, wind speed, radiation, and humidity on an hourly basis at the site since 2017. These data were used for AquaCrop model testing. In addition, the long-term weather data used in the model simulation were obtained from the China Meteorological Administration. The data include the daily maximum and minimum air temperature, wind speed, sunshine hours, relative humidity, and rainfall. The annual average precipitation in the wet year was 1084 mm, and 320 mm in the dry year.

Soils in the site were sampled at different depths for soil texture and soil water property measurements, as listed in

Table 1. Measured soil properties of the study area.

Parameter	Depth cm 0–20 cm	Depth cm 20–40 cm	Depth cm 40–60 cm
Clay	8.07%	10.46%	9.99%
Silt	43.26%	58.33%	57.42%
Sand	48.40%	31.20%	32.59%
Soil classification	loam	Silt loam	Silt loam
Field capacity	42.07 cm3cm−3	35.24 cm3cm−3	35.10 cm3cm−3
Wilting point	15.0 cm3cm−3	13.0 cm3cm−3	13.0 cm3cm−3
Saturated soil water content	51.69 cm3cm−3	44.59 cm3cm−3	44.55 cm3cm−3

The model default data was used for the missing land information.

. Rice and wheat rotation is generally practiced in the study area. Normally, winter wheat is not irrigated due to the high water table and appropriate rainfall during its growing season, generally from early November to early June. Supplemental irrigation for rice is needed from mid-June to mid-October, even though there is ample rainfall during the growing season. The recorded annual rice irrigation depth was about 900 mm in 2018 due to high evapotranspiration and low irrigation efficiency.

2.2. AquaCrop Model Description and Parameterization

AquaCrop is a mechanistic and dynamic crop growth model developed by the United Nations Food and Agriculture Organization (FAO). It contains four components: climate, crop, soil management, and field management. The model simulates crop growth and harvestable yields based on available water and regular field water. Both non-conservative and conservative crop parameters have been used in AquaCrop. Conservatively calculated parameters for the location, field management, temperature, and time were specified by crop form and cannot be used by default values directly. Non-conservative parameters, on the other hand, must be calibrated and evaluated before being included [7]. The Design AquaCrop is a program that simulates crop growth in reaction to accessible water and is used in the ongoing development of land, crop, and atmosphere systems. AquaCrop uses standard plant evapotranspiration (ETo) data to simulate soil evaporation components and crop transpiration separately, based on the daily canopy cover and soil drying. AquaCrop calculates the daily evapotranspiration with Equation (1) and biomass with Equation (2). Crop yield is obtained by multiplying biomass by the harvest index, as expressed in Equation (3). The adjustment of HI with the available water depends on the timing, severity, and duration of water stress [7,8].

$$\mathrm{ET}=\mathrm{E}+\mathrm{Tr}$$

(1)

$$\mathrm{B}=\mathrm{WP}{\displaystyle \sum \mathrm{Tr}}$$

(2)

$$\mathrm{Y}=\mathrm{B}\times \mathrm{HI}$$

(3)

where E is soil evaporation in mm, Tr is crop transpiration in mm, ET is evapotranspiration in mm, B is biomass yield in kg ha⁻¹, WP is the normalized crop water productivity in g m⁻², HI is harvest index, and Y is grain yield in kg ha⁻¹.

The average field WP is approximately constant for any particular climate and crop. For rice production, WP should be set between 15 and 20 gm⁻², the harvest index (HI) is between 35 and 50%.

The AquaCrop model inputs include the following:

(1)

Climate data: daily maximum and minimum air temperatures, (ETo) references, precipitation, and average annual carbon dioxide (CO₂) are required for the AquaCrop model. For this study, 64 years of historical weather data (1955–2018) for the area were obtained from the China Meteorological Administration, including air temperatures, wind speed, sunshine hours, and mean relative humidity (RH). The ETo with FAO-56 Penman–Montith equation was used with these weather data. Climatic data are displayed in Figure 1a,b.

(2)

Soil data: the number of characteristics of the soil layer, soil composition, saturated volumetric water, field capacity, permanent wilting point, and saturated hydraulic conductivity are used in model operations. The soil layer in this analysis has loamy to silt-loam soil, according to the “Malvern 3000 laser” test, and soil samples were taken at depths of 0 to 20, 20 to 40, and 40 to 60 cm from each region Table 1. There were no impermeable or restrictive soil textures at the test site, which would have hampered root development. Without measuring hydraulic soil properties, the model calculates the respective values based on soil texture. In the event that certain properties are calculated, users can input the measured attributes into the model.

(3)

Crop data: days of transplanting, maximum canopy cover, flowering, the beginning of senescence, maximum rooting, and physiological maturity. The crop input parameters needed by AquaCrop for rice and user input are listed in Table 2. Rice is a grain-producing crop classified as C3 category. On the 23 June 2018, the seedlings were transplanted (row spacing: plant spacing = 33 cm:12.0 cm). Crop growth was measured, including leaf area index (LAI), biomass accumulation, and rice yields. For LAI measurement, the specific gravity method was used (randomly selected from each plant). To do so, 20 leaves were taken and a 10 cm length was cut for the specific gravity leaf. The average width of each leaf was measured to calculate the LAI [9]. CC was determined by using the relationship between LAI and CC Equation (4) [7] as:

CC = 1.005(1 − exp (−0.6 × LAI))1.2

(4)

where CC is canopy cover, and LAI is the leaf area index.

Table 2. The crop input parameters for AquaCrop and the other parameters are not included left as the default values. The water productivity normalized for ETo and CO₂ concentration (WP^*).

Parameters	Value	Unit	Default Value
Maximum canopy cover (CCx)	98	%
Maximum effective root depth (Zx)	0.5	m	0.4–0.6
Recovery days after transplanting	5	days
Days from transplanting to max canopy	45	days
Days from transplanting to start senescence	77	days
Days from transplanting to maturity	110	days
Days of the flowering stage	11	days
Normalized Water productivity (WP*)	19	gm−2	19
Crop coefficient (Kctr)	1.2		1.1
Reference harvest index (HIo)	42	%	35–50

Table 2. The crop input parameters for AquaCrop and the other parameters are not included left as the default values. The water productivity normalized for ETo and CO₂ concentration (WP^*).

Parameters	Value	Unit	Default Value
Maximum canopy cover (CCx)	98	%
Maximum effective root depth (Zx)	0.5	m	0.4–0.6
Recovery days after transplanting	5	days
Days from transplanting to max canopy	45	days
Days from transplanting to start senescence	77	days
Days from transplanting to maturity	110	days
Days of the flowering stage	11	days
Normalized Water productivity (WP*)	19	gm−2	19
Crop coefficient (Kctr)	1.2		1.1
Reference harvest index (HIo)	42	%	35–50

The irrigation schedule was entered into AquaCrop by identifying the dates and depths of irrigation, with basin irrigation as the appropriate irrigation method. If there w still water on the soil surface or the daily rainfall is more than 20 mm, the model assumes irrigation delay. The field management was designed to resemble the expectedly good (non-stressing) bunds at the height of 25 cm. Surface runoff was assumed to be negligible due to surrounding bunds, controlled irrigation, and rainfall during the growing season.

2.3. AquaCrop Model Calibration

AquaCrop simulation includes soil and water balance, climate, and plant characteristics. It tests a daily water equilibrium between the soil, the cultivation, and the atmosphere, which leads to the production of biomass and food. It is concerned with the use of water in crop development as part of a water-driven model. Climate, crop, irrigation, farm management methods, and soil properties are the basic tools used to run the model. Model criteria in simulating measured grain yields were evaluated using Pearson’s coefficient of determination (R²), root mean square error (RMSE), normalized root mean square error (NRMSE), and index of agreement (d). The Pearson’s coefficient of determination (R²) ranges from 0 to 1 and values greater than 0.5 are considered suitable for watershed simulations. In NRMSE, simulation results can be regarded as excellent if NRMSE is less than 10%, good if it is between 10% and 20%, fair if it is between 20% and 30%, and bad if it is more than 30%. The value of the index of agreement d is between 0 and 1, where 0 indicates no agreement and 1 indicates a perfect agreement between the expected and the observed results [10].

$${\mathrm{R}}^2={\left[\frac{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}\left({\mathrm{o}}_{\mathrm{i}}-\overline{\mathrm{o}}\right)\left({\mathrm{p}}_{\mathrm{i}}-\left.\overline{\mathrm{p}}\right)\right.}}{\sqrt{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{o}}_{\mathrm{i}}-\overline{\mathrm{o}}\right)}^2}}\sqrt{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{p}}_{\mathrm{i}}-\overline{\mathrm{p}}\right)}^2}}}\right]}^2$$

(5)

$$\mathrm{RMSE}=\sqrt{\frac{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{o}}_{\mathrm{i}}-{\mathrm{p}}_{\mathrm{i}}\right)}^2}}{\mathrm{n}}}$$

(6)

$$\mathrm{d}=1-\frac{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{o}}_{\mathrm{i}}-{\mathrm{p}}_{\mathrm{i}}\right)}^2}}{{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{\left(\left|{\mathrm{p}}_{\mathrm{i}}-\overline{\mathrm{o}}\right|+\left|{\mathrm{o}}_{\mathrm{i}}-\overline{\mathrm{o}}\right|\right)}^2}}$$

(7)

where O_i is the ith observation, P_i is the ith predicted value, P is the mean of the predicted value, ō is the mean of the observed data, and n is the total number of observations [10].

Data from the Yanyun Irrigation District on-station experiment in Yangzhou in 2018 were used to calibrate the model, including CC, grain yield, and total aboveground biomass. To calibrate the system over the growing season, soil profile, texture, daily climate data from the China Meteorological Administration, field management practices, and irrigation regimes and schedules were used. Certain parameters that were not easily measured during the experiment were given default model values [8,11]. For paddy yield levels, the model parameters were changed by simulation using a trial and error method.

2.4. AquaCrop Model Application

For the water-saving mechanism, the crop’s water productivity and water use efficiency are determined to produce irrigation simulation scenarios for long-term period 1955–2014. Water productivity and water use efficiency (WUE) are important concepts in agricultural water management, and the AquaCrop model can be used to simulate and optimize these parameters for crop production. Water productivity is a measure of the amount of crop produced per unit of water used. It is usually expressed as kg/m³. In other words, it is the ratio of crop yield to applied water or evapotranspiration (ET). Water use efficiency (WUE) is another important parameter for agricultural water management. It is a measure of the amount of crop yield per unit of water transpired by the crop. WUE is usually expressed as kg/m³. In other words, it is the ratio of crop yield to total water used. As shown below, the grain yield/water ratio, or the grain yield/field (irrigation and rainfall) or irrigation ratio, could convey water productivity [12,13], as expressed in Equations (8) and (9). In order to investigate WUE under long-term weather conditions, simulations were performed using 60 years (1955–2014) of historical weather data. The required parameters for AquaCrop soil, crop, water management, and initial water conditions in this long-term simulation were based on the calibrated parameters in 2018, covering the lowest and highest water stress that might be experienced during the rice-growing season in the region. The irrigation quota’s scenario design should cover the minimum amount of irrigation that can be achieved in the area: rate and full irrigation quota. The soil moisture content was entered into the model as the percentage of RAW. Soil moisture content was designed to be 80%, 110%, 120%, 150%, 180%, 200%, and 220% of RAW, and simulated per irrigation water depths were 10, 20, 30, 40, 50, and 60 mm. A total of 42 irrigation scenarios were simulated for 60 years, and multi-objective and multi-scenario optimization of irrigation scheduling was conducted based on the AquaCrop modeling results. For the simulation period of 1955 to 2014, the simulation was started with a net irrigation requirement of 0% RAW, and different scenarios were simulated thereafter.

$$\mathrm{WPet}=\left[\frac{\mathrm{Y}}{\mathrm{ET}}\right]$$

(8)

$$\mathrm{IWP}=\frac{\mathrm{Y}}{\mathrm{I}}$$

(9)

where WPet is the normalized crop water productivity in kg m⁻³ and IWP is irrigation water productivity in kg m⁻³.

AquaCrop maintains a consistent water balance that includes infiltration, drainage, deep percolation, evaporation, and transpiration, as well as water content improvements [14]. In the experimental conditions, analyzing the water balance of rice fields, different rainfall years, and irrigation modes were taken into consideration. The Pearson type-III model method was used to conduct a hydrological frequency study, with three classes of wet, normal, and dry years corresponding to p ≤ 10%, 10% < p < 90%, and p ≥ 90%, respectively. The analysis identified the wet year as 1991, normal year as 1981, and dry year as 1966. For studying the change in rice yield with irrigation requirement, WUE, WPet, and IWP were calculated for the different hydrologic years.

3. Results and Discussions

3.1. AquaCrop Model Performance

The model accurately simulates the CC, in terms of calibration data. In certain cases, the simulated CC closely mirrored the observed CC as shown in Figure 2; the linear association between measured and simulated CC R² is 0.99 for 2018. In 2018, RMSE and NRMSE were 3.6% and 5.3%, respectively. Index of agreement d = 0.99 for 2018. Our results are consistent with those observed in the previous study by Xu 2019 [2]. The CC simulation error is lower than the majority of the AquaCrop model’s performance on different crops; for rice, Maniruzzaman 2015, Amiri 2014, Amiri 2016, Shrestha 2013, and Pirmoradian 2019; for soybeans, Paredes 2015; and for pea, Bello 2017 [4,11,15,16,17,18,19]. According to Jamieson 1991 [20], the efficiency of the simulation of CC rice was excellent. As a result, the AquaCrop model has a successful simulation of the rice canopy cover. In the current study, rice was grown in the drying season, which resulted in fewer tillers and lower total leaf area indexes than when rice was flooded, resulting in lower CC due to the water stress effects of rice leaf expansion. On the other hand, AquaCrop may lead to an overestimation of CC.

AquaCrop generated a good match during the plant’s growth stage between simulated and measured biomass as shown in Figure 3. RMSE, NRMSE, and d were 0.50 t ha⁻¹, 5.30%, and 0.99, respectively, in 2018. The variance between predicted and simulated biomass was observed by Paredes 2017, with errors of between −26% and 9.4%, respectively [21], and 2–10.5% by Xu [2]. The validated biomass NRMSE, between simulated and calculated biomass as recorded by Saadati 2011 and Maniruzzaman 2015, were from 11.2% to 12.4% above the current report, respectively [3,4]. In comparison, rice production was 8.511 t ha⁻¹ for 2018, overestimated by 2.3%. Our findings are similar to those reported in Shao [22] and were met by Xu 2019; 6.351 t ha⁻¹, 6.557 t ha⁻¹ for 2012–2013 [2]. In comparison with Raoufi, Amiri, Amiri, and Ebrahim [11,15,23,24], these experiments revealed that the model was capable of accurately simulating the production and yield of these plants. While a wide range of plants has been studied in various temperature and agro-environmental conditions, rice, a staple food that offers essential nutrition to more than half of the world’s population, has received little attention. China, India, Tanzania, Bangladesh, Iran, West Africa, and Southeast Asia have all conducted rice research utilizing AquaCrop.

The AquaCrop model outputs also include simulated daily soil water content (SWC). In AquaCrop, the total available water (TAW) is defined as water between field capacity and the wilting point. When soil moisture content is less than the field capacity, water stress existed, which is acceptable for upland crops, but in comparison to upland crops, rice is more sensitive to water stress, giving a different TAW value.

Soil water storage dynamics are presented during the 2018 season in Figure 4. Soil moisture characteristics (field capacity and wilting point) are also presented in Figure 4. It is clear that AquaCrop accurately predicted both the amount and trend of soil moisture content. Even so, the observed soil moisture was slightly overestimated most of the time. The simulated moisture content was generally higher than the field capacity, except in the late season, when water content started to drop below the field capacity. Flooding water depth, field drain spacing, and boundary condition also played important roles in percolation, yet these factors were ignored by AquaCrop. There should be some changes in the procedure of drainage. Water depth, boundary condition, and the field drain spacing when there was pipe drainage system in field should be taken into account in the procedure of percolation. During the 2018 monitoring period, the value of RMSE and NRMSE for soil moisture evaluation were 11.2 mm, and 3.20%, respectively, and, in addition, to the acceptable values of R² 0.60 and d 0.96, showed good simulation accuracy. Andarzian 2011 reported that the measured RRMSE of wheat soil moisture content was 0.035 for full irrigation and 0.04 for water deficit irrigation [25]. The model performances were all higher than 0.696, and others were close to 0.805, indicating a fair simulation accuracy. The mathematical study of Dirk Raes’ soil water balance research resulted in an EF of 0.21 for Mornag and an EF of 0.83 for the Kou Valley [26]. Abdalhi 2020 reported that simulated and measured SWC could be approved statistically by the low values of NRMSE = 7.6% and RMSE = 15.6%, with acceptable values in R² and d [27]. However, overall patterns in average magnitude and relatively different models simulated and calculated SWC are defined as overestimates based on the majority of the point distributions in graphs, according to the model, which appears to overestimate SWC simulations [7,17,21,27,28,29,30] seemed to be understated by AquaCrop in other experimental studies with different opinions.

3.2. Predicted Water Balance of Paddy Fields under Different Irrigation Strategies

The improved water system procedure enables the current and forecasted soil moisture to be kept between two soil moisture thresholds: The percolating soil flux begins to be substantial (field capacity) with higher relative soil moisture, whereas the crop begins to suffer because of soil water shortage (crop stress) with lower relative soil moisture. Furthermore, this standard backing the proper planning of the water system and the measure of water for every water system permits one to lessen the entries over the field limit edge, decreasing permeation motion and preserving water system volume while evapotranspiration remains almost steady.

When compared to current irrigation practices, the optimized irrigation strategy allows one to increase irrigation efficiency and water productivity, saving a significant percentage of water, fertilizer, and energy. Figure 5a presents Tmax and Tmin, and Figure 5b presents rainfall and ETo for the period between 1955 and 2014. Based on the calibration results in 2018, to identify optimal irrigation schedules, the optimal irrigation schedule model was developed using AquaCrop in multiple irrigation scenarios over several design years. Figure 6a presents the predicted relationship between the rice yield and net irrigation requirement for the period of 1955–2014. Figure 6b presents the changes of WP, IWP, and WUE with the net irrigation for the same period of time.

For the selected different years, Figure 7 presents the predicted yields under the different treatment conditions for three years: 1991, 1981, and 1966, with the wet year as 1991, the normal year as 1981, and the dry year as 1966. There are two categories for the rice yield change with the irrigation requirements. The first category is a fast rice increase process, where rice yield increases rapidly as irrigation demand increases, and the second category is the gradual stabilization stage. If irrigation continues to increase, rice yields are not significantly increased, but are either higher or fluctuate slightly. On the other hand, when the irrigation continues to increase, the rice yield will not increase dramatically, but will either remain higher or fluctuate slightly. The study has looked at significant values for the rainy year compared to the dry year because the yield was higher in the wet year. Similar results were reported by Shao 2018, Wang 2018, and Zhai 2019 [22,31,32].

Table 3. Compared to the productivity of the rice crop.

Treatment	Rainfall mm	ETo mm	Irrigation Depth mm	% of Water Saved	Yield t/ha	Reduction in Yield%
wet year 1991
Net Irr	1084	462	370		8.53
Irr 80% RAW			160	57%	8.52	0.09
Irr 110% RAW			120	68%	8.51	0.21
Irr 120% RAW			80	78%	8.4	1.5
Irr 150% RAW			40	89%	7.91	7.25
normal year 1981
Net Irr	564	493	386		8.28
Irr 80% RAW			200	48%	8.2	0.94
Irr 110% RAW			180	53%	7.98	3.62
Irr 120% RAW			160	59%	7.86	5.01
Irr150% RAW			100	74%	6.8	17.9
dry year 1966
Net Irr	320	531	500		7.59
Irr 80% RAW			260	48%	7.24	4.59
Irr 110% RAW			220	56%	6.5	14.37
Irr 120% RAW			200	60%	6.5	14.3
Irr150% RAW			120	76%	4.98	34.33

presents irrigation depth and percentage of water saved with deficit irrigation scenarios for 1991, 1981, and 1966, as well as yield and reduction in yield. One year from each year group was chosen for further analysis. Table 3 displays the results of these analyses. Compared with the dry year 1966, the rainfall amount in the wet year 1991 was nearly four times larger, and irrigation was reduced by approximately 30%. Most of the rainfall was drained away without being fully utilized. As a result, throughout the season, the surface water depth could be replenished in a timely manner.

Table 3 also shows the water-saving effects of different irrigation practices. For the wet year 1991, when irrigation was 120% RAW, the percentage of water saved was 78% and reduction in yield was 1.5%, compared with 150% RAW, where the percentage of water saved was 89%. The reduction in yield was 7.25% in 1991. For the normal year 1981, when irrigation was 80% RAW, the percentage of water saved was 48% and reduction in was yield 0.98%; with 120% RAW, the percentage of water saved was 59%. The reduction in yield was 5.01% in 1981. For the dry year 1966, when irrigation was 80% RAW, the percentage of water saved was 48% and reduction in yield was 4.59%; this is compared with 110% RAW, where the percentage of water saved was 56%. The reduction in yield was 14.37% in 1966. As a result, in wet years, saved water has less effect on yield until irrigation of 120% RAW. In normal years, irrigation of 120% RAW to saved water could be used. For dry years, yield was affected by saved water.

WUE and water productivity (WPet, IWP) for the three years selected in the scenario of irrigation were 80%, 110%, 120%, and 150% RAW with irrigation quota, 40 mm for wet, and 20 mm for normal and dry years. As shown in Figure 8, water productivity parameters were used to measure the irrigation strategy in the rainfall comparison years. In the rainy year, IWP (irrigation water productivity) was substantially higher, and the amount of water used was much lower. On the other hand, WPet was more extensive in the wet year because the yield was higher, and the ET was smaller than in the dry year. Finally, the WPet and IWP increased again in the rainy year because the yield was higher, but the WUE was lower.

The relationship between rice yield and irrigation in the three different years were defined by fitting based on the changing trend of the yield with the irrigation, as shown in Figure 8a. The fitted functional relations, where y1, y2, and y3 represent the fitted relationships between rice yield and irrigation water quantity in 1991, 1981, and 1966, give a maximum value of R² as 0.99 in a normal year, R² 0.98 in a dry year, and R² 0.86 in a wet year, which indicates the acceptability of the fitting accuracy. Figure 8b shows that the highest WPet in the wet year, 1.77kg m⁻³, was obtained for 80% and 120% RAW in year 1991, while the lowest WPet in the dry year 1966 was 1.13 kg m⁻³ for 150% RAW. As shown in Figure 8c, the highest IWP was 19.78 kg m⁻³ was obtained for 150% RAW in the wet year and 9.32 kg m⁻³ in the normal year, while the lowest IWP in the dry year, 1.90 kg m⁻³, was for the net irrigation requirement. As shown in Figure 8d, the highest WUE of 1.84 kg m⁻³ was obtained for 80%, 110%, and 120% RAW in the dry year, while the lowest WUE in the wet year, 0.59 kg m⁻³, was for the net irrigation requirement.

Table 4. WUE, WPet, and IWP for different crops.

Crop	Exp. Year	WUE	WPet	IWP	References
rice	2018	in net irrigation 2.06 to 2.51 kg m−3 in deficit irrigation 1.58 to 2.57 kg m−3			[33]
	2020		for continuous flooding in humid 0.82 kg m−3 in dry 0.76 kg m−3		[34]
	2012		1.66 g kg−1	1.81 g kg−1	[35]
	2002	0.80 g kg−1			[36]
	2003	0.36 g kg−1			[36]
	2001	2.2 g kg−1	in continuously flooded 0.2–0.4 g kg−1 in India and 0.3–1.1 g kg−1 in Philippines		[12]
pea	2017	0.52–1.08 kg m−3	0.83–1.54 kg m−3	1.12–9.4 kg m−3	[21]
wheat	2020	0.58–0.66 kg m−3	1.4–1.5 kg m−3		[37]
	2008		0.65–1.72 kg m−3		[38]
	2002		1.13 kg m−3		[39]
	2009		0.39–0.65 kg m−3		[40]
	2011		In normal year 1.37–1.62 kg m−3 in dry year 0.47–1.73 kg m−3		[25]
	2014	0.7–1.3 kg m−3			[41]
maize	2019	7.69–20.57 kg m−3			[42]

shows some of the values of WUE, WPet, and IWP for different crops.

4. Conclusions

In this study, the AquaCrop model was used for simulating rice irrigation strategies in the Yanyun irrigation area on the lower reaches of the Yangtze River Basin, China. The results showed that effective irrigation planning is needed to increase water usage and productivity. In wet years, water saving has less effect on yield until deficit irrigation at 120% RAW. In normal years, applying irrigation at 120% RAW may save water. In dry years, yield was affected by irrigation water reduction. The use of an irrigation scheduling model is required to help farmers make decisions in real-time. It is desirable that future research should provide a full economic review and an evaluation of the various degrees of water stress on the efficiency of the rice yield. This has shown that an optimized irrigation schedule can ensure a high rice yield and water saving in the irrigation area. Future research may include economic analysis to evaluate the costs and benefits of different irrigation strategies, including the costs of irrigation infrastructure, labor, water use, and the potential benefits in terms of crop yield and water productivity. This can help to identify the most cost-effective irrigation management strategies.