Simulation of Soil Water Changing and Ryegrass Growth in the Hilly Regions Based on the Epic Model
Wenping Zhou1, Muslim Qadir1, 2, Xuechun Wang1*, Shaikh Abdullah1, 2, Guotao Yang1
School of Life Science and Engineering & College of Adult and Online Education, Southwest University of Science and Technology, Mianyang 621010,China.
Faculty of Agriculture & Water Resources management, Lasbela University of Agriculture, Water and Marine sciences (LUAWMS) Lasbela, Baluchistan, 74200-Pakistan
ABSTRACT: To clarify the dynamic changing of soil water has great meaning for the improvement of water management in ryegrass land, to evaluate EPIC model is better to improve and accurate the EPIC model for ryegrass land, also it will promote the application of cropping system simulation and decision making technology in forage-grain cropping system. By field experimental and simulating method, this paper researched the growth process of ryegrass and dynamic changing of soil water in ryegrass land and clarified the accuracy of EPIC (Environmental Policy Integrated Climate) model to simulate ryegrass growth and soil water changing. Results showed 1). Though available soil water was relative lower during March to May, it recovered better during May to October in deep soil (0.4-0.6 m). Reasonable increase fertilizer for ryegrass land was better to increase forage yield and leaf area index at the hilly region of Sichuan Province. 2) Correlation index between observed and simulated available soil water in 0-1 m soil was 0.86-0.95, with the RRMSE value of 6.0%-17.2%; correlation index between observed and simulated soil water distribution in 0-1m soil was 0.57-0.92; most of them was higher than 0.85. 3) Correlation index between simulated and observed value of forage yield, leaf area index and height of ryegrass were all higher than 0.90, there is not significant difference between simulated and observed value for forage yield, leaf area index and height of ryegrass. The EPIC model can simulate soil water and forage yield of ryegrass land well, it can be used as a better tool to evaluate and research the suitableness of grain-forage cropping system at the hilly regions of Sichuan Province.
Key words: Ryegrass; EPIC model; Forage yield; Soil water; Nitrogen fertilizer
Ryegrass is a cool-season grass. Because it’s nutrient-rich (crude protein content of more than 25%) (Song et al., 2014), high digestibility (Liu et al.,2009), good palatability, strong cold resistance and other characteristics (Zhai et al., 2013 ; Chen, 2015), it is widely planted in the world. It is one of the most important cold season forages in hilly region of China. Annual rainfall in the hilly areas of Sichuan is 800-1000 mm, annual average temperature is 16-18°C, and the frost-free period is 230-340 d (Liang et al., 2016). It is suitable for the growth of forage crops such as ryegrass from a climatic point of view, but seasonal drought and soil erosion are serious. Water stress which limits the adaptability of ryegrass in Sichuan hilly areas is still controversial. Statistics show that precipitation in Sichuan hilly areas accounts for more than 70% of the annual precipitation from June to September(Pang et al., 2010), but due to high precipitation intensity and high surface runoff (Annual average soil erosion module is 3798 to 9831 t/(km2•a), the proportion of precipitation on-site infiltration is low. The total amount of precipitation from November to April is low. Without irrigation, deep soil moisture becomes the key to high yield of ryegrass. In the process of the introduction of alfalfa (Medicago sativa) in the Loess Plateau, it was discovered that the forage crops, such as alfalfa, have serious soil water depletion. After a few years of cultivation, the soil layer will easily form a permanent dry layer (Luo et al., 2015; Fang et al., 2010). It is not conducive to the sustainable use of soil water. Therefore, it is necessary to clarify the use of ryegrass in deep soil water in the hilly region of Sichuan, and then to improve the water management level of ryegrass grassland.
Constructing a reasonable grass-grain rotation system can effectively improve soil water use efficiency(Melake et al., 2011; Yang et al., 2015; Yang, 2015). The growth cycle of ryegrass is similar to that of winter wheat (Triticum aestivuml) ( Liu et al., 2015 ; Hao et al.,2015) and can theoretically replace the winter wheat in the traditional rotation system (Wheat-maize / soybean). Whether or not a new planting pattern (rye grass-corn/sweet potato) can be widely used, it is necessary to determine whether it is suitable for the environmental conditions in the hilly areas of Sichuan and whether it is conducive to the sustainable use of soil water. The combination of field experiments and crop model simulations is an effective method and hot spot for assessing the adaptability between crop production systems and the environment. EPIC model (Environment Policy Integrated Climate) is a multi-crop general model constructed by Williams et al in 1984 (Williams et al., 1984). It includes more than ten submodules such as meteorology, soil, and crop growth. Through the cooperation among various sub-modules, the model can perform long-term quantitative simulation of crop growth, water cycle, etc. in complex crop production systems. It can be used to evaluate the adaptability of crop production systems to the environment and is an effective tool for agricultural system decision-making (Williams, 1990; He and Cai, 2016; Williams, 1995). Most research results confirmed the simulation accuracy of crop yield and soil moisture of winter wheat, spring corn, and alfalfa using the EPIC model. All model users agree that before using the EPIC model for simulation studies, it is necessary to comprehensive evaluation of simulation accuracy (Wei et al., 2014; Xu et al., 2016; Camargo and Kemanian, 2016). At present, there is no report on the verification of the application of the EPIC model in the ryegrass grass field in Sichuan Basin area. Whether the EPIC model can be applied to the simulation of ryegrass and whether other modules need further improvement during the simulation is not yet clear.
Based on the observed data of artificial rye grassland from 2011 to 2015, this study analyzed the changes of soil water in ryegrass grassland, defined the use of ryegrass for deep soil moisture, and provided necessary for soil moisture management in ryegrass grassland in Sichuan hilly areas. Based on the comparison between the field observations and the simulated values of the EPIC model, the simulation accuracy of the EPIC model on the growth process and soil water dynamics of the ryegrass was analyzed. This laid the foundation for the application of the EPIC model in the crop rotation system in Sichuan hilly areas and provides a new approach and method for establishment of a regional grass-grain rotation system.
2. MATERIAL AND METHODS
2.1 Study Area
Mianyang is located in the northwestern part of Sichuan Basin and the middle and upper reaches of Fujiang River. There are a wide range of hills and flat valleys in Mianyang, with an average elevation of 362 m. It is one of the typical hilly areas in Sichuan Basin. The average annual precipitation is 826-1417 mm and the average annual temperature is 14.7-17.3 ?, with an average annual sunshine duration of 1300-1328 h and an average frost-free period of 253-301 d, which belongs to the humid monsoon climate in the subtropical zone.
The experiment was carried out in hills and mountains of Qingyi Town in Mianyang from 2011 to 2015. The soil of the experimental field was yellow soil with a soil bulk density of 1.1 g / cm, perennial moisture content of 9.2% -20.1%, wilting point of 6.2% -8.5% and organic matter content of 0.23 g / kg, total Nitrogen, total Phosphorus and total Potassium were 1.1, 0.22 and 15.6 g/kg, respectively.
2.2 Field experiment
2.2.1 Experiment design
Four fertility treatments designed for this experiment as followed, CK, no nitrogen fertilizer used; N1, applying 75 kg/hm2 of pure nitrogen; N2, 150 kg/hm2 of pure nitrogen; and N3, 225 kg/hm2 of pure nitrogen. With a randomized block design, 3 repeated treatments in 12 plots with the plot area of 20 m2 (4 m × 5 m). A 1 m wide interval between plots designed for sample taking and field management convenient. An annual ryegrass is planted in mid-September of each year with a seeding rate of 15 kg/hm2 and a row spacing of 0.15 m. In early January, early February, early March and early April of the following year, rye grass harvested. After the first three mowing cycles, nitrogen fertilizer was topdressing according to designed dosage. The method of nitrogen fertilizer operation was 50% basal fertilizer and 50% dressing (16.6% each time). Field management such as sowing and fertilizing are same with the local management.
2.2.2 Sample Collection
Before planting ryegrass, soil samples were collected in 10 cm layers from the soil surface to a depth of 1 m soil, and then to determine soil water content for each layer by oven dry method. After planting ryegrass, soil water content is measured once a month. Each plot is divided into two parts. One parts used to measure the dynamic changing of leaf area index, each 15 days measured one time, from which no harvest for rye grass from planting to die. Another part used to investigate the dynamic changing of soil water and the forage yield of ryegrass. In this part rye grass harvested in 1 m2 in a suitable time to measure the forage yield, and the soil sample were taken each month to measure soil water content each month.
After the last harvest, corn was planted. Field management practices such as sowing and fertilization of maize are consistent with high-yielding management in the area. Inside, five plant height (the height from the ground to the top of the ryegrass in the state of natural growth) were selected and averaged. Leaf area index was measured by canopy analyzer (TOP1300). Soil moisture measured in 2 fixed points by TDR (Time-Domain Reflectometry) probe which were buried in soil before experiment started. According to soil bulk density, soil volumetric water content is converted into weight water content. The conversion formula is ? = ?w, where ? is soil density, ? is volumetric water content, and w is mass water content. The soil moisture content, effective soil water content, plant height and leaf area index of each treatment were averaged over three replicates for statistical analysis.
Available soil water content (ASW) is calculated by equation 1.
* MERGEFORMAT (1)
Where ASW is 0-1 m available soil water in mm; n is the maximum number of soil layers measured; SWi is the soil water content in weight (%) of the soil layer i; WPi is the soil wilting humidity (%) of the soil layer i; Pi is the soil bulk density (g/cm3) , For the first i layer thickness in cm (Wang et al., 2004).
Simulation accuracy of ryegrass growth and soil water dynamics of EPIC model was evaluated by the following formula.
* MERGEFORMAT?2?* MERGEFORMAT?3?
* MERGEFORMAT?4?* MERGEFORMAT?5?
* MERGEFORMAT?6?* MERGEFORMAT?7?
Where RMSE is the root mean square error; RRMSE is the relative root mean square error; ME is the model efficiency; RE is the relative error; R is the correlation coefficient; R2 is the decision coefficient; Si is the simulated value; Mi is the observed value; * MERGEFORMATis the average; * MERGEFORMAT is the simulated average (Williams, 1995).
3. RESULTS AND ANALYSIS
3.1 Simulation of Soil Water Dynamics in Rye Grassland by EPIC Model
3.1.1 Monthly variation of available soil water in rye grassland
Fig.1 Comparison of available soil water of different nitrogen fertilization level in ryegrass land of the hilly regions at Sichuan Province
Under different levels of nitrogen fertilizer, available soil water of 0-1 m soil layer in rye grassland varied with the monthly precipitation change from 2011 to 2015 (Fig. 1). From April to September, mean monthly precipitation was 120 mm, which was significantly higher than the mean monthly precipitation (67.5 mm) from January to December, therefore available soil water rye grassland tended to increase during April to September. From October to May, mean monthly precipitation was 14.2 mm, which was significantly lower than the mean monthly precipitation (67.5 mm) from January to December, so available soil water in rye grassland decreased during October to May.
Overall, mean available soil water in 0-1 m soil layer was 125mm in ryegrass land during March to May, which was the lowest during the growing period of rye grass. Comparing with other ryegrass production periods, soil water content in ryegrass land was higher from July to September, with mean available soil water of 201 mm in 0-1 m soil. Considering the effective use of soil water, appropriate irrigation during February to May will be beneficial to improve the forage yield of ryegrass in Sichuan Basin area.
From 2011 to 2015, the soil available water in 0-1 m soil layer under CK, N1, N2 and N3 treatments were 176, 166, 168 and 151 mm, respectively. The simulation results of soil available water content are slightly lower than the observed values; The RRMSE value was lower and the difference between ME and R2 was smaller (Figure 2). There was no significant difference between the simulated and observed values of 0-1 m soil available soil water (P; 0.05). It shows that the dynamic changes of available soil water in the rye grassland simulated by EPIC model are acceptable in the statistical range.
3.1.2 Dynamic changing of soil water content among different soil layer in ryegrass land
From 2011 to 2015, with the increase of soil depth, the fluctuation of soil moisture showed a decreasing trend (Figure 3). Soil moisture in 0-0.2 m soil layer fluctuated greatly; with a difference of 16% between the lowest soil moisture content (6%) and the highest (22%). Soil moisture in the soil layer of 0.8-1.0 m was relatively stable, and the difference between the lowest value (8%) and the highest value (12%) of soil moisture content was 4%. From January to December, the soil moisture content showed a trend of first increasing and then decreasing.
Soil water in deep layer (0.4-0.6 m) increased significantly with the increase of monthly precipitation from May to October, and fluctuated from November to April (Fig. 3). This result showed that ryegrass had a certain depletion of soil water in the month of low precipitation. Soil water content of 0-0.2 m, 0.3-0.6 m, 0.8-1.0 m soil layer in the ryegrass field were 6%-10%, 7%-9% and 8%-9%, respectively during January to April period. From May to October, the soil water of corresponding soil layers were 8%-20%, 8%-15% and 8%-13% respectively. This result showed that soil water in deep soil gradually recovered in May-October from January to April.
Comparison results of simulated and observed soil water distribution in different soil layers (Fig.4) showed that the EPIC model is more reasonable to simulate the distribution of soil moisture profiles in dryland rye grassland of Sichuan hilly region. The R value between simulated and observed values of soil water distribution was 0.57-0.92, of which, 82.3% was greater than 0.85, 10.2% was less than 0.62; RMSE was 0.34%-2.10%, less than 1% accounted for 86.1% , More than 2% accounted for 1%.
From 2011 to 2015, the values of R between simulated and observed values of soil water content in 0-1 m soil layer were 0.78, 0.88 and 0.67 respectively in January, May and September (Fig. 4), with the RMSE values of 0.72 %, 0.63% and 1.01%, the RRMSE values of 8.56%, 6.24% and 7.76%, and the ME values of 0.88, 0.72 and 0.81, respectively. In September with more precipitation, the R value between the simulated value and the observed value is lower, and the difference between R2 and ME values is larger, indicating that the accuracy of the EPIC model to simulate soil moisture profile is lower in the drought months.
Under the treatments of CK, N1, N2 and N3, the simulated values of soil water content in 0-1 m soil layer were 15.1%, 16.0%, 15.0% and 14.5%, respectively, decreased by 9.5% and 8.9% compared with observed value. Its R value is 0.73-0.88, R2 value is 0.62-0.78, ME value is 0.75-0.81, RMSE value is 0.45%-0.72% and RRMSE value is 5.21%-8.69%, RE value is -9.5% ~ 9.4%. The values of RRMSE less than 10% and the differences between ME and R2 were less than 0.15, which indicated that the EPIC model simulated the soil water content of ryegrass with different nitrogen treatments reasonably and could reflect the variation law of soil moisture under different treatments. At the same time, the R, R2 and ME values of CK and N1 treatments were significantly higher than those of N2 and N3 treatments (Table 1), indicating that the simulation accuracy of soil moisture under CK and N1 treatments was higher than that of N2 and N3 treatments.
Table 1 Comparison of simulated and observed soil water content by weight under different nitrogen treatments during 2011-2015 in ryegrass land of the hilly regions in Sichuan Province
Fertilizer treatment R R2 ME RMSE (%) RRMSE (%) RE (%)
CK 0.88 0.78 0.81 0.45 5.21 -9.5
N1 0.79 0.75 0.80 0.64 5.64 8.9
N2 0.73 0.65 0.75 0.61 6.48 -8.2
N3 0.74 0.62 0.78 0.72 8.69 9.4
3.2 Growth process of ryegrass and its simulation
3.2.1 Plant height of Ryegrass
From 2011 to 2015, the mean height of ryegrass was 73-85 cm. during 2011-2012 production years the plant height of all the nitrogenous fertilizer treatments was not significantly different (P ;0.05); during 2012 to 2013 period, the plant height of N2 and N3 treatments was significantly higher than that of CK, but there was no significant difference in the plant height between N1 and CK treatments; during 2014 to 2015 production years, the plant height of N2 and N3 treatments was significantly higher than that of CK, but the plant height of N2 and N3 treatments was significantly higher than that of N1. It shows that the planting of ryegrass on the yellow soil in the hilly area of Sichuan requires a certain amount of nitrogen fertilizer to ensure its normal growth and continuous high yield.
The simulated values of plant height of ryegrass with CK, N1, N2 and N3 treatments were 79.2, 80.5, 81.6 and 85.2 cm, respectively, 1.0% higher, 2.0% higher, 0.5% lower and 2.0% higher than the observed value. The R values between the simulated and observed values of the plant height were greater than 0.90, the ME values were greater than 0.95, and the difference between the simulated values and the R2 values was less than 0.05 (figure 5), indicating that the differences between plant height simulated values and observed values of ryegrass were small, ryegrass plant height simulated by EPIC model was accurate.
In different years, the plant height of ryegrass simulated by EPIC model was 1.85% and 2.71% lower than the observed values in 2011-2012 and 2014-2015 respectively, but higher than the observed values in 2012-2013 and 2013-2014 6.55% and 4.51%, respectively (Table 2). The values of R, RRMSE and ME between simulated and observed values were lower in 2012-2013 and 2013-2014. The RMSE between simulated and observed values were in 2012 – Higher in 2013 and 2013-2014. It shows that the simulation accuracy of the plant height of the ryegrass in the year with more precipitation (2012-2013, 2013-2014) is lower in the EPIC model than in the years with less precipitation (2011-2012, 2014-2015).
Table 2 Comparison of simulated and observed high of ryegrass at the hilly regions of Sichuan Province
(%) R RMSE
(%) R2 ME
2011-2012 79.5 81.0 -1.85 0.98 1.5 3.4 0.90 0.85
2012-2013 84.6 79.4 6.55 0.92 2.3 3.5 0.87 0.80
2013-2014 83.5 79.9 4.51 0.95 2.6 4.1 0.88 0.81
2014-2015 80.5 82.8 -2.71 0.97 1.1 3.1 0.91 0.88
3.2.2 Dynamic variation of leaf area index
Nitrogen fertilizer had a significant effect on leaf area index of ryegrass (Table 3). From 2011 to 2015, the average maximum leaf area index of ryegrass under CK, N1, N2 and N3 treatments were 2.43, 2.63, 3.33 and 3.40, respectively, of which N1, N2 and N3 treatments were significantly higher than those of CK, N2 and N3 treatments were significantly higher At N1. The results showed that proper nitrogen fertilizer increased leaf area index of ryegrass. Affected by precipitation, the average maximum leaf area index of ryegrass in 2011-2012 and 2014-2015 was 3.0 and 3.1, respectively, significantly higher than the corresponding values (2.8 and 2.9) in 2012-2013 and 2013-2014.
Table 3 Comparison of simulated and observed leaf area index of ryegrass at the hilly regions of Sichuan Province
Fertilizer treatment 2011-2012 2012-2013 2013-2014 2014-2015
CK 2.5a 2.3a 2.4a 2.5a
N1 2.6a 2.5b 2.6b 2.8b
N2 3.4b 3.2c 3.1c 3.6c
N3 3.5b 3.2c 3.3d 3.6c
Significant differences (p;0.05) between different treatments were indicated by different letters.
The average simulated values of maximum leaf area index of ryegrass under CK, N1, N2 and N3 treatments were 2.3, 2.7, 3.3 and 3.5 cm, respectively. The R values between the simulated and observed values of LAI were both greater than 0.90, the ME values above 0.80, and the differences between R2 and ME values were less than 0.05 (Figure 6), indicating that within the statistical range. The simulation of the dynamic change of leaf area index of ryegrass with EPIC model was accurate.
Compared with the observed values, the simulated values of leaf area index of ryegrass decreased 3.45% in 2013-2014, 0.01% in 2012-2013, 3.33% and 3.23% respectively in 2011-2012 and 2014-2015 % (Table 4). The values of R, RRMSE and ME between simulated and observed values were lower in 2012-2013 and 2013-2014. The RMSE between simulated and observed values were lower in 2012-2013 and higher in 2013-2014 (Table 4). It shows that the simulation accuracy of the leaf area index of the ryegrass in the year with more precipitation (2012-2013, 2013-2014) is lower than that in the years with less precipitation (2011-2012, 2014-2015). Although the difference of LAI was relatively large in 2011-2014, Analysis of variance showed that there was no significant difference between the simulated and observed values (P; 0.05), indicating that the simulation accuracy of EPIC model on ryegrass leaf area index statistically acceptable.
Table 4 Comparison of simulated and observed leaf area index of ryegrass at different years in the hilly regions of Sichuan Province
Year Simulated Observed RE (%) R RMSE(cm) RRMSE (%) R2 ME
2011-2012 3.1 3.0 3.33 0.95 0.08 5.4 0.83 0.85
2012-2013 2.7 2.7 0.01 0.89 0.11 7.5 0.77 0.80
2013-2014 2.8 2.9 -3.45 0.91 0.21 6.1 0.75 0.79
2014-2015 3.2 3.1 3.23 0.93 0.09 4.1 0.81 0.80
3.3 Forage yield of ryegrass land
The annual grass production of ryegrass was 36.7, 51.9, 65.0 and 70.9 t / hm2, respectively, in CK, N1, N2 and N3 from 2011 to 2015. Compared with CK, the yields of N1, N2 and N3 increased 41.4 %, 77.1% and 93.1%. Analysis of variance showed that the yield of N1, N2 and N3 was significantly higher than CK; the yield of N2 and N3 was significantly higher than that of N1; and the difference between N2 and N3 was not significant (Table 5). The results showed that the application of nitrogen fertilizer could significantly increase the yield of ryegrass. Under the circumstance of cutting 4 times, the nitrogen application rate of artificial rye grassland in hilly area of Sichuan Province should not be lower than that of N1 (75 kg / m2 of pure nitrogen) and does not exceed the level of N3 (225 kg/hm2of pure nitrogen).
Table 5 Comparison of simulated and observed ryegrass yield under different nitrogen treatments at the hilly regions of Sichuan Province during 2011-2015
Fertilizer treatment 2011-2012 2012-2013 2013-2014 2014-2015 Average
CK 37.5a 35.2a 35.4a 38.5a 36.7
N1 53.6b 50.2b 51.1b 52.8b 51.9
N2 67.5c 62.1c 61.8c 68.7c 65.0
N3 71.5c 70.1d 68.3c 73.6c 70.9
Simulated annual forage yield of ryegrass under CK, N1, N2 and N3 treatments from 2011 to 2015 were 36.5, 52.0, 65.0 and 71.0t/hm2 respectively, which was similar with observed forage yield. The R values between the simulated and observed values were all above 0.90 and the ME values above 0.95 (Figure 7), indicating that the difference between the simulated and observed values of ryegrass was small. The simulation of ryegrass forage yield with the EPIC model was more accurate.
From year to year, except from 2011 to 2012, the simulated values of ryegrass production in 2012-2015 are lower than the observed values. Among them, the simulated production of ryegrass in 2012-2013, 2013-2014 and 2014-2015 is lower than the observed output .Lower by 0.37%, 0.18% and 0.34% respectively. The R’s value, RRMSE’s value and ME’s value between simulation value and observation value are lower in 2012-2013 and 2013-2014. RMSE value between simulation value and observation was higher in 2012-2013 and 2013-2014 (Table 6). It indicates that the accuracy of simulation of the yield of ryegrass in the year with more precipitation (2012-2013, 2013-2014) is lower than that of the years with less rainfall (2011-2012, 2014-2015).
Table 6 Comparison of simulated and observed ryegrass yield in different years at the hilly regions of Sichuan Province
Year Simulated yield (t/hm2) Observed yield (t/hm2) RE
(%) R RMSE (t/hm2) RRMSE
(%) R2 ME
2011-2012 58.0 57.5 0.87 0.95 0.21 0.39 0.96 0.98
2012-2013 54.2 54.4 -0.37 0.93 0.48 0.84 0.97 0.96
2013-2014 54.1 54.2 -0.18 0.91 0.51 0.87 0.95 0.94
2014-2015 58.2 58.4 -0.34 0.98 0.23 0.43 0.96 0.97
4.1 Dynamic changing of soil water in rye grassland
Studies by Sun et al (2014) showed that the coverage of the ground with ryegrass had a certain influence on the infiltration of precipitation, and the cumulative infiltration of precipitation showed an increasing trend with the coverage of ryegrass. Although there was some ryegrass cover in the surface of the field from March to May in this study, a large amount of deep soil water was consumed due to the relatively low average monthly rainfall (only 13.2 mm). Soil available water content decreased significantly at 1 m soil layer (Figure 1). Related research shows that the roots of ryegrass are mainly concentrated in 0 – 0.3 m soil layer (Wang et al., 2014; Xie et al., 1981), the actual average monthly precipitation in July-September is significantly higher than the annual average monthly precipitation. In addition, therefore, the soil moisture of ryegrass increased significantly at this stage, especially in deep (0.4-0.6 m) soil (Figure 3).Affected by precipitation and crop consumption, the soil available water content of 0-1 m soil layer in rye grassland was 125 mm in March-May, which was significantly lower than the annual average. The distribution of soil water profile showed that Jan to April, the grassland of ryegrass had a certain consumption of deep soil moisture, and deep soil moisture could be completely recovered from May to October, which is beneficial to the sustainable utilization of soil moisture in the rye grassland. From the perspective of sustainable use of soil moisture, it is suitable to grow ryegrass in Sichuan hilly area.
Williams et al (1995) improved the simulation module for groundwater level in EPIC model in 1995, Baier-Robertson introduced in 1998 (Cavero et al.,1988) and Green and Ampt infiltration equation introduced in 2000 (Lui et al., 1993). These improvements and improvements, significantly improved the simulation accuracy of the EPIC model for the water cycle in crop production systems. Li et al (2015) and Hao et al (2015) think that the EPIC model can effectively simulate the dynamic changes of soil moisture. The study shows that the R value of effective water content of 0-1 m soil layer simulated by EPIC model is 0.86-0.95 and the RRMSE value is 6.0%-17.2%. The monthly variation of soil water simulated by EPIC model is similar to the actual the agreement between observations is high, which is similar to that of Li Jun et al (2010) in the Loess Plateau. At the same time, this study also shows that the simulation accuracy of the soil moisture content of the ryegrass field is slightly lower in the richer months of the EPIC model, and the focus should be on improving and applying the model in the future.
4.2 Ryegrass growth process and forage yield
Nitrogen fertilization is one of the effective ways to increase forage yield. A large number of production practices and fertilizer experiments have demonstrated that nitrogen fertilizer can significantly improve forage quality (Zhou, et al., 2010; Li et al., 2011). The results of Zhan et al (2011) showed that the tiller numbers, plant height and leaf area of ryegrass increased significantly under NPK treatments, compared with those without nitrogen fertilizer. Results of this paper showed that with the prolongation of experimental time, the effect of nitrogen fertilizer on the plant height of ryegrass gradually appeared. From 2011 to 2012, the plant height of ryegrass under four treatments was not significantly different. From 2014 to 2015, plant height of N1, N2, and N3 were all significantly higher than that of CK. The plant height of N2 and N3 treatments was significantly higher than that of N1. From 2011 to2015, the average maximum leaf area index of N1, N2 and N3 was significantly higher than that of CK, N2 and N3 treatments were significantly higher N1, increase nitrogen supply, is conducive to increasing leaf area index of ryegrass.
Huang et al. (2010) showed that the ryegrass yield was significantly increased with the increase of nitrogen application level in the rye grassland of Fujian Province. The yield of ryegrass was the highest at the nitrogen level of 200 kg/hm2 Increase the input of nitrogen fertilizer, rye grass nitrate content increased, is not conducive to the health of livestock and poultry. This study shows that the appropriate fertilization in the hilly areas of Sichuan can significantly increase the yield of ryegrass forage grass, and the yield of all the fertilized ryegrass is higher than that of the non-nitrogenous fertilizers, which is similar to previous studies. In addition, the yield of ryegrass with N1 treatment (nitrogen fertilizer application rate of 75 kg/hm2) was not significantly different from that of CK (no nitrogen fertilizer application) when harvesting for 4 stubs. The N3 treatment (225 kg/hm2 fertilizer application rate) the difference between wheat straw yield and N2 treatment (200 kg/hm2) was not significant. Therefore, the amount of nitrogen fertilizer (pure nitrogen) in the artificial rye grassland in the hilly area of Sichuan should not be less than 75 kg/hm2 and not more than 225 kg/hm2.
Some papers reported that simulation results of EPIC model can simulate the average yield well over a long period, but cannot reflect the annual variation of crop production among different years (Williams et al., 1990; He and Cai, 2016; Williams,1995; Xu et al., 2016). The EPIC model is considered to be an effective tool to calculate the impact of environment on agricultural production over a long period (Raniero et al., 2014; Wei et al., 2014; Camargo et al., 2016). Results of this paper showed that R value between simulated and observed ryegrass leaf area were greater than 0.90, and ME values were all greater than 0.80. The R values were greater than 0.90 between simulated and observed ryegrass forage yield, and the ME values were greater than 0.95. This result indicated that the EPIC model can simulate well not only the leaf area index but also the plant height and the forage yield; it can be used in ryegrass simulation study.
Soil moisture in ryegrass field was lower during March to May period, and soil moisture in deep layer (0.4-0.6 m) recovered well from May to October in the hilly area of Sichuan Province. Appropriately increasing nitrogen application could significantly improve the growth of ryegrass. The maximum nitrogen used for ryegrass was not less than 75 kg/hm2 and not more than 225 kg/hm2 under the conditions that the ryegrass harvest four times.
The R value between the simulated and observed available soil water in 0-1 m soil layer ranged from 0.86 to 0.95, and its RRMSE values ranged from 6.0% to 17.2%. The R value between simulated and observed soil water dynamic changing ranged from 0.57 to 0.92. Overall, the dynamic changes of soil moisture in the ryegrass field simulated by EPIC model are more accurate and can be used to study the dynamic changes of soil moisture in the ryegrass field.
The R value was greater than 0.90, either between simulated and observed plant height or between simulated and observed leaf area index. This result indicated that the EPIC simulated leaf area index and plant height of ryegrass well. The R value between simulated and observed values of ryegrass yield was greater than 0.90, and the difference between the ME and R2 values was less than 0.02, these results indicated that EPIC model can simulate ryegrass forage yield well.