Response of Alfalfa Yield to Rates and Ratios of N, P, and K Fertilizer in Arid and Semi-Arid Regions of China Based on Meta-Analysis

Abstract

Quantifying the impacts and contributions of nitrogen (N), phosphorus (P), and potassium (K) fertilizer management on the annual and cutting cycle yields of alfalfa can provide guidance for alfalfa cultivation. In this study, 597 sets of alfalfa yield data from the arid and semi-arid regions of China were collected and grouped according to the N, P, K, and NPK rates. Statistical and meta-analyses were employed to explore the response of alfalfa yield to fertilization management. The results indicated that the annual and cutting cycle yields of alfalfa increased and then decreased as the N rate increased; meanwhile, the annual yield increased with the P and K rates, and the cutting cycle yield decreased with increased cutting cycles under P and K treatments. The alfalfa yield at the first cutting cycle was about 30% of the annual yield in the arid and semi-arid regions of China. Moreover, the annual yield increased and then decreased as the NPK rate increased. The Aggregated Boosted Tree (ABT) algorithm-based analysis showed that the NPK rate had the greatest contribution to the annual yield (37.61%), followed by the application rates of P (24.50%), N (22.55%), and K (15.34%). The impacts of the P/NPK, N/NPK, and K/NPK ratios on the annual yield were 38.64%, 31.71%, and 27.65%, respectively. Additionally, an NPK amount > 225–310 kg/ha and ratios of N, P, and K of 14.28–27.72%, 36.36–50%, and below 25%, respectively, resulted in the highest increase in alfalfa yield. This is recommended as the optimal fertilization practice to obtain high alfalfa yields and improve fertilizer use efficiency in the arid and semi-arid regions of China.

1. Introduction

Alfalfa (Medicago sativa L.) is hailed as “the king of forage grass” and “feed queen”, and is one of the most widely distributed and cultivated forage crops in China and globally [1,2,3]. China is the second largest alfalfa-growing country in the world after the United States, with its cultivated area accounting for 15% of the global cultivation area [4]. The arid and semi-arid regions, with an annual precipitation below 400 mm, are the main alfalfa production areas in China. The provinces of Gansu, Inner Mongolia, Ningxia, and Shaanxi, located in the arid and semi-arid regions, account for about 80% and over 60% of China’s cultivated alfalfa area and production, respectively [5]. To ensure the development of animal husbandry and the effective supply of meat and milk, China imports a large amount of alfalfa hay each year, with imports increasing from 18,000 tons in 2008 to 1.78 million tons in 2022 [6], accounting for about a quarter of the global imports. In recent years, the cultivated area for alfalfa in China has expanded from 433,300 hectares in 2018 to 546,700 hectares in 2020, with a growth of 26.20%; correspondingly, the alfalfa yields also increased from 3.77 million tons to 4 million tons, a growth of 6.10% [5]. However, the significant gap between the alfalfa supply and demand has not been fundamentally resolved, making increasing alfalfa production in China an urgent need [7].

In the arid and semi-arid regions of China, alfalfa cultivation still largely follows traditional methods [8]; that is, no fertilizer or only one type of fertilizer (e.g., phosphorus) is applied. These fertilization practices have prevented alfalfa from reaching its full production potential. Scientifically formulating fertilization strategies is currently a crucial issue that needs to be addressed to increase the yield of alfalfa in these regions [9]. The reasonable application of nitrogen (N), phosphorus (P), and potassium (K) fertilization could effectively increase the yield and quality of alfalfa, while delaying senescence and maintaining high production performance [10,11,12].

Compared with no N application, a higher annual alfalfa yield was produced under an N application rate of 75–120 kg/ha in Xinjiang [13,14]. An N application rate of 42–52.50 kg/ha was recommended for high alfalfa yield in the Gansu and Xinjiang provinces [8]. Compared with no P application, in a field with a soil available phosphorus level below 10 mg/kg, the annual yield of alfalfa increased by 37.30% under a P application rate of 150 kg/ha [15]. Applied 100–120 kg/ha of P resulted in an annual yield of alfalfa that was 15.40–39.96% higher than that with no P application [16,17]. The annual yield of alfalfa increased by 500 kg/ha under a P application rate of 292.50 kg/ha [18]. Applied 120 kg/ha of P resulted in an annual alfalfa yield that was 5.72% higher than that with no P application [19]. Compared with no K application, the annual alfalfa yield increased by 4.81% under a K application rate of 100 kg/ha; the alfalfa yield in each cutting cycle and annual yield showed an initial increase and then decreased as the K application rate increased [20].

The total amount and ratio of N, P, and K all affected the yield, regenerative capacity, and economic benefits of alfalfa [15]. However, we cannot rely solely on the individual application rates of N, P, or K as the basis for fertilization management practices to increase alfalfa yields [9]. The total NPK amount also impacts the alfalfa yield. Compared with no application, the annual yield of alfalfa increased by 26.85% under an NPK application rate of 232.50 kg/ha [21]. An NPK application level of 201.75–298.50 kg/ha increased the annual yield of alfalfa by 46.40–48.09%, compared with no fertilization [22,23]. The under an N:P:K also affects alfalfa yield. The highest annual yield of alfalfa was achieved under N:P:K ratio of 1:6:36, representing a 28.89% increase over no fertilization [24]. Higher annual alfalfa yields were achieved under N:P:K ratios of 1:2:1 and 1:2:2 [23].

To date, the studies on the effects of N, P, and K fertilizer on alfalfa yield in the arid and semi-arid regions of China is primarily based on a limited number of studies in different regions [13,14,15,16,17,18,19,20,21,22,23,24,25,26,27]. The fertilization measures adopted to increase alfalfa yields differ across regions, leading to notable differences and even contradictions in the conclusions regarding the effects of fertilization on alfalfa yield [8,13,14]. For instance, due to the nitrogen-fixing ability of the roots, there has been a long-standing controversy over whether nitrogen fertilizer should be applied for alfalfa cultivation [8]. This demonstrates the complexity of the impact of N, P, and K amounts and ratios on alfalfa yield and introduces uncertainty into fertilization management practices for alfalfa crops. It is also difficult to quantitatively analyze the effects of fertilization on alfalfa yield based solely on the scattered experimental results from specific locations [28]. To the best of our knowledge, few attempts have been made to comprehensively quantify the impacts of fertilizer management on alfalfa yield in the arid and semi-arid regions of China.

Filling this knowledge gap is vital to increasing alfalfa yields and optimizing fertilizer management practices in the arid and semi-arid regions of China [29]. In this study, meta-analysis was employed to quantify the impact and contributions of N, P, K, and the NPK rate and ratios on alfalfa yield, and appropriate fertilization measures are proposed to maximize alfalfa yields by synthesizing the previous studies in the arid and semi-arid regions of China.

2. Materials and Methods

2.1. Data Sources and Classification

Relevant studies and data on the impact of fertilization on alfalfa yield since 2000 were retrieved from the Google Scholar, Web of Science (WOS), CNKI, and WanFang databases. The following keywords were used: “alfalfa yield”, “fertilization”, “fertilizer”, “N application rate”, “P₂O₅ application rate”, and “K₂O application rate”. To enhance the representativeness of the obtained data, the retrieved studies were further screened according to the following criteria: (1) the experimental sites were located in the arid and semi-arid regions of China; (2) the experiments consisted of field trials with control groups and used different fertilization treatments (different N, P₂O₅, K₂O, and NPK application rates and ratios); (3) the yield of alfalfa at different cutting cycles was recorded, including textual descriptions, charts, and so on, with experimental durations of ≥2 years or with ≥2 sets of data for the same experiment; (4) the number of replicates for each treatment was 3 or higher; and (5) articles with duplicated experimental contents were excluded. After this screening, 597 sets of alfalfa yield observation data were selected.

The statistical characteristics of the different fertilization treatments are shown in

Table 1. Statistical characteristics of different fertilization treatments in the literature.

Category	Min	Max	Mean	SD	25th Quantile	50th Quantile	75th Quantile
N	30.00	277.52	103.53	59.59	60.00	90.00	138.00
P	21.18	877.50	138.50	127.93	60.00	105.00	180.00
K	20.00	540.00	139.20	110.11	60.00	100.00	180.00
NPK	86.70	1334.25	386.70	263.94	225.00	310.00	447.30

Note: Min—annual minimum value; Max—annual maximum value; Mean—annual mean value; SD—standard deviation. The unit of all data is kg/ha.

. The number of samples and alfalfa yields under different fertilization treatments are listed in

Table 2. The number of alfalfa yield data points under different annual fertilizer application rates.

Category	N (kg/ha)					P (kg/ha)
	0	>0–50	>50–100	>100–150	>150	0	>0–60	>60–120	>120–180	>180
Yield number	41	23	40	22	20	78	61	66	38	47
Category	K (kg/ha)
	0		>0–60		>60–120		>120–180		>180
Yield number	38		34		40		17		32

and Figure 1. The non-fertilized group was considered the control group. The yield results were grouped using the 25th and 75th quantiles as the cutoffs. The other grouping thresholds were mainly determined based on the average value and the 50th quantile. For simplicity, N, P, K, and NPK refer to the annual pure application rates of N, P₂O₅, K₂O, and N + P₂O₅ + K₂O in the different fertilization treatments, respectively.

2.2. Meta-Analysis

To determine whether the results have statistically significant differences, it is necessary to test the heterogeneity of the sample data (Q-test). The formula is as follows [8]:

$$Q={\displaystyle {\sum }_{i=1}^K{w}_i(\ln R_i}{)}^2-{\displaystyle \frac{({\displaystyle {\sum }_{i=1}^K{w}_i\ln R_i{)}^2}}{{\displaystyle {\sum }_{i=1}^K{w}_i}}}$$

(1)

when P_Q > 0.05 (chi-squared test), it indicates that the data do not exhibit heterogeneity, and a fixed-effects model should be chosen; conversely, if P_Q < 0.05, it suggests that the data have heterogeneity, and a random-effects model should be used.

The effect size (lnR) was used to evaluate the impact of different fertilization treatments on alfalfa yield. lnR > 0 indicates that the fertilization treatments have a positive effect on alfalfa yield; conversely, lnR < 0 it indicates a negative effect. A larger lnR value suggests a more pronounced effect. The formula for lnR is as follows [8]:

$$\ln R=\ln ({\displaystyle \frac{X_T}{X_C}})$$

(2)

where X_T and X_C represent the average alfalfa yields of the treatment group and the control group, respectively. The formula for the variance (V) of the lnR is as follows [28]:

$$V={\displaystyle \frac{S{D}_T^2}{n_T{x}_T^2}}+{\displaystyle \frac{S{D}_C^2}{n_C{x}_C^2}}$$

(3)

where SD_T and n_T represent the standard deviation and number of replicates for the treatment group, respectively, and SD_C and n_C represent the standard deviation and number of replicates for the control group, respectively. The overall or average (combined) effect size lnR₊₊ for the treatment group is obtained based on the weight and sum of different data pairs; the formula is as follows [8]:

$$\ln R_{++}={\displaystyle \frac{{\displaystyle \sum (w_i\times \ln R_i)}}{{\displaystyle \sum w_i}}}$$

(4)

where lnR₊₊ represents the weighted combined effect size, and lnR_i and w_i are the effect and weight of the ith pair of data, respectively. The weight of each pair of data is the reciprocal of the variance of the corresponding effect size; the formula is as follows [8]:

$$w={\displaystyle \frac{1}{V}}$$

(5)

The 95% confidence interval (95%CI) for the weighted combined effect size lnR₊₊ was calculated as follows:

$$95\%CI = ln{R}_{++}\pm 1.96S_{\ln R_{++}}$$

(6)

where S_InR++ represents the standard deviation of the weighted combined effect size lnR₊₊. With the zero point of the horizontal axis as the boundary, if the 95%CI includes 0, it indicates that the effect of the different fertilization treatments on alfalfa yield was not significant. If the 95%CI is entirely located to the right or left of 0, it indicates that different fertilization treatments had a significant promoting or inhibiting effect on alfalfa yield, respectively (p < 0.05) [30].

To describe the magnitude of the impact of fertilization treatments on alfalfa yield, the lnR₊₊ was converted into the percentage change in the yield (E) relative to the control group [31].

$$E=[EXP(\ln R_{++})-1]\times 100\%$$

(7)

2.3. Data Processing

The GetData 2.2 software was used to extract data from the literature. The Excel 2016 software was used for data merging and classification. The MetaWin 2.0 software was used to calculate the effect size and 95%CI for each group, and the random-effects model was selected based on the results of the heterogeneity test. Data visualization was performed using the SigmaPlot 14.0 and Origin 2018 software. The importance of the explanatory variables was assessed using Aggregated boosted Tree (ABT) [32], analysis and the “xgboost” package in Spyder (Python 3.11), in order to quantitatively evaluate the relative effects of fertilization on alfalfa yield.

3. Results

3.1. Effect of N Application Rate on Alfalfa Yield

3.1.1. Effect of N Application Rate on Annual Yield of Alfalfa

The mean annual N application rate for alfalfa fields in the arid and semi-arid regions of China was 103.53 kg/ha (Table 1), and the 25th and 75th quantiles were 60.00 and 138.00 kg/ha, respectively.

The statistical results for the annual alfalfa yield under different N application rates in the arid and semi-arid regions of China are shown in

Table 3. Statistical results for annual alfalfa yield under different fertilization treatments.

Category	AR (kg/ha)	Min	Max	Mean	SD	CV	25th Quantile	50th Quantile	75th Quantile
N	0	2776	22,011	10,588	5353	0.51	7374	9658	12,357
	>0–50	3458	31,415	13,738	8316	0.60	10,878	14,188	16,534
	>50–100	3813	34,017	14,162	9056	0.64	7881	12,823	15,025
	>100–150	3966	18,409	12,096	4619	0.38	8390	12,152	14,550
	>150	3715	15,033	10,784	4626	0.43	7770	9230	14,814
P	0	1562	23,018	10,806	5089	0.47	8202	10,987	13,810
	>0–60	2802	32,416	12,484	5175	0.41	7800	12,688	14,808
	>60–120	3287	32,516	13,082	6981	0.53	8390	13,139	15,655
	>120–180	3216	35,017	13,755	9290	0.68	5729	12,726	16,708
	>180	4248	17,765	14,028	3804	0.27	12,110	14,311	16,508
K	0	2776	16,608	10,276	4365	0.42	8202	9311	13,070
	>0–60	2934	33,616	11,342	7714	0.68	9061	9871	10,890
	>60–120	3813	34,017	12,184	7399	0.61	9230	10,149	14,933
	>120–180	5574	26,563	13,196	8007	0.61	8905	12,106	12,833
	>180	7344	27,013	13,422	5495	0.41	9483	13,678	14,814

Note: AR—application rate. The unit of all data is kg/ha.

. The annual yield of alfalfa ranged from 2776 to 34,017 kg/ha under the different nitrogen application rates. The mean annual alfalfa yields were 10,588, 13,738, 14,162, 12,096, and 10,784 kg/ha under N application levels of 0, >0–50, >50–100, >100–150, and >150 kg/ha, respectively. The coefficient of variation (CV) of the annual yield of alfalfa was at the medium degree of variation under the different treatments.

According to Table 3, the mean annual yield of alfalfa increased and then decreased as the N application rate increased. The mean annual yield of alfalfa was the highest (14,162 kg/ha) and the CV was the largest (0.64) under an N application rate of >50–100 kg/ha. Compared with no N application, the mean annual yield of alfalfa increased by 29.75%, 33.76%, 14.24%, and 1.85% under N application levels of >0–50, >50–100, >100–150, and >150 kg/ha, respectively.

3.1.2. Effect of N Application Rate on Alfalfa Yield at Different Cutting Cycles

As shown in Figure 2, the average alfalfa yield of each cutting cycle increased and then decreased as the N application rate raised. Compared with no N application, the average yield of the first cutting cycle increased by 32.71% and 30.89% under N application rates of >0–50 and >50–100 kg/ha, respectively. There was no significant difference in the average yield of the first cutting cycle under an N application rate of >100–150 kg/ha, while it decreased by 28.62% under an N application rate of >150 kg/ha compared with no N application. Additionally, the yields at the first, third, and fourth cutting cycles all showed decreasing trends as the N application rate increased. Compared with an N application rate of >0–50 kg/ha, the average yields at the first cutting cycle decreased by 1.37%, 24.67%, and 46.22% under N application rates of >50–100, >100–150, and >150 kg/ha, respectively. The alfalfa yield at the second cutting cycle increased and then decreased as the N application rate increased. Compared with an N application rate of >0–50 kg/ha, the average yields of the second cutting cycles increased by 8.44%, 1.20%, and 2.58% under N application rates of >50–100, >100–150, and >150 kg/ha, respectively.

Moreover, the average alfalfa yield at the first cutting cycle accounted for 30.93%, 28.81%, 30.40%, 30.85%, and 23.60% of the annual yield under the different N application rates. Thus, the alfalfa yield at the first cutting cycle accounted for about 30% of the annual yield in the arid and semi-arid regions of China.

As shown in Figure 2, the alfalfa yield decreased as the cutting cycle increased under the N application rate of less than 100 kg/ha. However, the yield at the second cutting cycle was the highest under N application rates exceeding 100 kg/ha. The fitting formulas for the average yield of alfalfa (kg/ha) and cutting cycles (x, values of 1, 2, 3, and 4) were as follows (Figure 3):

y_N1 = 5133.7e^−0.182x R² = 0.9502

y_N2 = 5136.3x^−0.11 R² = 0.2960

y_N3 = 6064.9e^−0.131x R² = 0.9157

y_N4 = −482x² + 1521.4x + 3142.50 R² = 0.9798

y_N5 = −998.75x² + 4414.1x − 435.75 R² = 0.9930

where y_N1, y_N2, y_N3, y_N4, and y_N5 represent the average alfalfa yields at N application rates of 0, >0–50, >50–100, >100–150, and >150 kg/ha, respectively.

3.1.3. Effect of N Application Rate on Percentage Change in Alfalfa Yield

Compared with no nitrogen application (Figure 4a), the annual percentage change in alfalfa yield significantly increased (p < 0.05) under N application rates of >0–50, >50–100, and >100–150 kg/ha, while the annual percentage change in alfalfa yield increased (but not significantly) under an N application rate over 150 kg/ha. The annual percentage change in alfalfa yield increased and then decreased as the N application rate increased. The annual percentage change in alfalfa yield increased by 19.72%, 22.15%, 19.77%, and 6.78% under N application rates of >0–50, >50–100, >100–150, and >150 kg/ha, respectively.

Compared with the yield at each cutting cycle without N application (Figure 4b), the yield at the first and second cutting cycles significantly increased (p < 0.05) by 16.73% and 24.50% and the yield at the third cutting cycle increased by 18.80% under an N application rate of >0–50 kg/ha. Similarly, the yields at the first and second cutting cycles significantly increased (p < 0.05) by 19.31% and 11.38% and the yields at the third and fourth cutting cycles increased by 12.65% and 25.79% under an N application rate of >50–100 kg/ha. The yields at the first, second, and third cutting cycles increased by 15.87%, 31.08%, and 11.37% under an N application rate of >100–150 kg/ha, but only the effect at the second cutting cycle was significant (p < 0.05). Under an N application rate of >150 kg/ha, the percentage change in the first and third cutting cycle yields decreased by 3.25% and 4.43%, and the second cutting yield increased by 7.60%. As the N application rate increased, the percentage change in each cutting cycle yield showed a trend of initially increasing and then decreasing.

3.2. Effect of P Application Rate on Alfalfa Yield

3.2.1. Effect of P Application Rate on Annual Yield of Alfalfa

The average annual P application rate for alfalfa fields in the arid and semi-arid regions of China was 138.50 kg/ha (Table 2), and the 25th and 75th quantiles were 60.00 and 180.00 kg/ha, respectively.

The annual yield of alfalfa ranged from 1562 to 35,017 kg/ha under different P application rates (Table 2). The mean annual alfalfa yields were 10,806, 12,484, 13,082, 13,755, and 14,028 kg/ha under P application levels of 0, >0–60, >60–120, >120–180, and >180 kg/ha, respectively. The CV of the annual yield of alfalfa was at the medium degree under the different treatments.

According to Table 3, the average annual yield of alfalfa increased as the P application rate increased. The mean annual yield of alfalfa was the highest (14,028 kg/ha) and the CV was the smallest (0.27) under a P application rate of >180 kg/ha. Compared with no P application, the mean annual yield of alfalfa increased by 15.53%, 21.06%, 27.29%, and 29.82% under P application levels of >0–60, >60–120, >120–180, and >180 kg/ha, respectively. This indicates the average annual yield of alfalfa in the arid and semi-arid regions of China was sensitive to P application.

3.2.2. Effect of P Application Rate on Alfalfa Yield at Different Cutting Cycles

As illustrated in Figure 5, the average yields of alfalfa at each cutting cycle decreased as the cutting cycle increased under different P application rates. Moreover, the average yields of alfalfa at the first, second, and third cutting cycles without P were all lower than those with P application. Compared to without P application, the average yields at the first cutting cycle increased by 18.78%, 16.89%, 24.19%, and 32.97% with P application rates of >0–60, >60–120, >120–180, and >180 kg/ha, respectively. The yield changes at the second and third cutting cycles of alfalfa with P application were similar to those of the first cutting cycle. Compared to without P, the average yield at the fourth cutting cycle increased by 3.12% and 2.23% under P application rates of >0–60 and >60–120 kg/ha, while the average yields decreased by 2.76% and 17.84% under P application rates of >120–180 and >180 kg/ha, respectively. Additionally, the alfalfa yields at the first and second cutting cycles decreased and then increased as the P application rate increased. Compared with a P application rate of >0–60 kg/ha, the average yield at the first cutting cycle decreased by 1.59% under a P application rate of >60–120 kg/ha, while it increased by 4.55% and 11.95% under P application rates of >120–180 and >180 kg/ha, respectively. The yield of alfalfa at the third cutting cycle increased and then decreased as the P application rate increased. Compared with a P application rate of >0–60 kg/ha, the average yield at the third cutting cycle increased by 1.87% under a P application rate of >60–120 kg/ha, while it decreased by 3.02% and 22.41% under P application rates of >120–180 and >180 kg/ha, respectively. The yield of alfalfa at the fourth cutting cycle decreased as the P application rate increased. Compared with the average yield of alfalfa at the fourth cutting cycle under a P application rate of >0–60 kg/ha, the average yields decreased by 0.86%, 5.7%, and 20.32% under P application rates of >60–120, >120–180, and >180 kg/ha, respectively.

Under different P application rates, the yield of the first cutting cycle of alfalfa accounted for 29.78%, 29.53%, 29.25%, 30.93%, and 34.47% of the annual yield, respectively. It was concluded that the alfalfa yield of the first cutting cycle accounted for about 30% of the annual yield under P application.

In addition, the yield of alfalfa under different P application rates decreased as the cutting cycles increased (Figure 6). The fitting formulas for the average alfalfa yield (kg/ha) and cutting cycles (x, values of 1, 2, 3, 4) were as follows:

y_P1 = 4938.3e^−0.142x R² = 0.9354

y_P2 = 6342.4e^−0.173x R² = 0.8367

y_P3 = 6211.2e^−0.167x R² = 0.7967

y_P4 = 6980.9e^−0.211x R² = 0.8613

y_P5 = 8156.5e^−0.305x R² = 0.9263

where y_P1, y_P2, y_P3, y_P4, and y_P5 represent the average yields of alfalfa under P application rates of 0, >0–60, >60–120, >120–180, and >180 kg/ha, respectively.

3.2.3. Effect of P Application Rate on Percentage Change in Alfalfa Yield

Compared with no P application (Figure 7a), the annual percentage change in alfalfa yield significantly increased (p < 0.05) with P application. The annual percentage change in alfalfa yield was 18.79%, 29.42%, 33.48%, and 37.09% with P application rates of >0–60, >60–120, >120–180, and >180 kg/ha, respectively. Compared to without P (Figure 7b), the alfalfa yields under an application rate of >0–60 kg/ha at the first, second, and third cutting cycles significantly increased (p < 0.05) by 15.31%, 20.66%, and 18.03%, and the yield of the fourth cutting cycle also increased by 34.78%; however, this difference was not significant. The percentage change trends for alfalfa yield at the different cutting cycles under P application rates of >60–120, >120–180, and >180 kg/ha were similar to those under a P application rates of >0–60 kg/ha.

3.3. Effect of K Application Rate on Alfalfa Yield

3.3.1. Effect of K Application Rate on Annual Alfalfa Yield

The average annual K application rate of alfalfa fields in the arid and semi-arid regions of China was 139.20 kg/ha, and the 25th and 75th quantiles were 60.00 and 180.00 kg/ha, respectively (Table 2).

The average annual alfalfa yield was 10,276 kg/ha (Table 3) without K application. The average annual alfalfa yield was 11,342, 12,184, 13,196, and 13,422 kg/ha under K application rates of >0–60, >60–120, >120–180, and >180 kg/ha, respectively.

According to Table 3 the average annual yield of alfalfa increased as the K application rate increased. The average annual alfalfa yield (13,422 kg/ha) was the highest and the CV (0.41) was the smallest under a K application rate of >180 kg/ha. Compared with the average annual yield of alfalfa without K, the average annual yield of alfalfa increased by 10.37%, 18.57%, 28.42%, and 30.61% under K application rates of >0–60, >60–120, >120–180, and >180 kg/ha, respectively.

3.3.2. Effect of K Application Rate on Alfalfa Yield at Different Cutting Cycles

As illustrated in Figure 8, the first and second cutting cycle alfalfa yields increased as the K application rate rose. The average yields at the first cutting cycle increased by 31.05%, 33.32%, 34.90%, and 37.41% with K application rates of >0–60, >60–120, >120–180, and >180 kg/ha, respectively, compared to that without K application. The yield at the third cutting cycle increased and then decreased as the K application rate increased compared with that without K application. The average yields of the third cutting cycle with increased by 13.89%, 5.74%, and 2.66% under K application rates of >0–60, >60–120, and >120–180 kg/ha, respectively, while the average yield with a K application rate of >180 kg/ha decreased by 5.56%. There were fewer yield data points (2, 2, 1, and 2, respectively) for the fourth cutting cycle under the different K application rates, making it difficult to discern a trend. Additionally, the alfalfa yields of the first and second cutting cycles increased as the K application rate increased. Compared with a K application rate of >0–60 kg/ha, the average yield at the first cutting cycle increased by 1.73%, 2.94%, and 4.85% with K application rates of >60–120, >120–180, and >180 kg/ha, respectively. The alfalfa yield of the third cutting cycle decreased as the K application rate increased. Compared with a K application rate of >0–60 kg/ha, the average yields of the third cutting cycle with K application rates of >60–120, >120–180, and >180 kg/ha decreased by 7.16%, 9.86%, and 17.08%, respectively.

The yield at the first cutting cycle of alfalfa accounted for 28.21%, 29.79%, 31.47%, 30.37%, and 36.35% of the annual yield under the different K application rates, respectively. The alfalfa yield at the first cutting cycle accounted for about 30% of the annual yield under different K application rates.

The yield of alfalfa under the different K application rates decreased as the cutting cycle increased (Figure 9). The fitting formulas for the average alfalfa yield (kg/ha) and cutting cycle (x, values of 1, 2, 3, 4) were as follows:

y_K1 = 4454.6e^−0.123x R² = 0.8403

y_K2 = 5206.1e^−0.097x R² = 0.9070

y_K3 = 5693.7e^−0.151x R² = 0.9837

y_K4 = 4778.9x^−0.189 R² = 0.5546

y_K5 = 9988.4e^−0.468x R² = 0.8202

where y_K1, y_K2, y_K3, y_K4, and y_K5 represent the average yields of alfalfa with K application rates of 0, >0–60, >60–120, >120–180, and >180 kg/ha, respectively.

3.3.3. Effect of K Application Rate on Percentage Change in Alfalfa Yield

Compared with no K application (Figure 10a), the annual percentage change in alfalfa yield significantly increased (p < 0.05) under K application. The annual percentage change in alfalfa yield with K application rates of >0–60, >60–120, >120–180, and >180 kg/ha were 5.82%, 11.97%, 16.34%, and 19.85%, respectively. Compared with the alfalfa yield of each cutting cycle without K application (Figure 10b), the yield percentage change in each cutting cycle of alfalfa under the different K application rates showed a decreased trend as the cutting cycle increased.

The alfalfa yields under a K application rate of >0–60 kg/ha increased by 13.29% (p < 0.05) and 6.13% in the first and second cutting cycles, but the yield at the third cutting cycle decreased by 5.39%. The yields increased by 15.46% (p < 0.05), 12.96% (p < 0.05), and 4.01% for the first, second, and third cutting cycles under a K application rate of >60–120 kg/ha. The percentage change in yield under a K application rate of >60–120 kg/ha was 49.18% and 8.68% for the first and second cutting cycles, but the percentage change in the third cutting cycle decreased by 14.45%. The yield increased by 20.26%, 12.17% (p < 0.05), 12.70%, and 8.30% in the first, second, third, and fourth cutting cycles under a K application rate of >180 kg/ha. Overall, K application had the strongest effect on the yield at the first cutting cycle of alfalfa.

3.4. Effects of NPK Application Amount and Ratio on Alfalfa Yield

A total of 86 sets of NPK fertilization combination data were collected in this study. The 25th, 50th, and 75th quantiles for yield with the application of NPK and different application ratios of N, P, and K were calculated. The 25th, 50th, and 75th quantiles were used as the thresholds for classifying the annual fertilization amounts and application ratios of N, P, and K into four levels [33], which were further categorized into four categories (

Table 4. Classification of four types of annual fertilization amounts and application ratios of N, P, and K.

Category	Meaning	Level (Code)	Lower Limit (kg/ha)	Upper Limit (kg/ha)
NPK (kg/ha)	N + P + K	Low (T1)	0	225
		Medium–low (T2)	225	310
		Medium–high (T3)	310	447.30
		High (T4)	447.30	—
N Application Ratio (%)	N/NPK	Low (N1)	0	14.28
		Medium–low (N2)	14.28	27.72
		Medium–high (N3)	27.72	39.48
		High (N4)	39.48	100
P Application Ratio (%)	P/NPK	Low (P1)	0	18.63
		Medium–low (P2)	18.63	36.36
		Medium–high (P3)	36.36	50
		High (P4)	50	100
K Application Ratio (%)	K/NPK	Low (K1)	0	25
		Medium–low (K2)	25	36.36
		Medium–high (K3)	36.36	49.17
		High (K4)	49.17	100

Note: low, medium–low, medium–high, and high levels represent the 0th–25th, 25th–50th, 50th–75th, and 75th–100th quantiles, respectively.

).

The average annual NPK application rate of alfalfa fields in the arid and semi-arid regions of China was 386.70 kg/ha, and the 25th and 75th quantiles were 225.00 and 447.30 kg/ha, respectively (Table 1).

The annual percentage change in alfalfa yield showed a significant increase and then decrease with increasing NPK application rate (p < 0.05) (Figure 11). The annual percentage change in alfalfa yield reached the maximum value (41.35%) at the T2 level and the minimum value (30.39%) at the T3 level.

As the N application ratio increased, the annual percentage change in alfalfa yield showed a significant increase and then decrease (p < 0.05). The annual percentage change in alfalfa yield reached the maximum value (42.51%) at the N2 level and the minimum value (23.17%) at the N4 level. Similarly, as the P application ratio increased, the annual percentage change in alfalfa yield exhibited a significant increase and then decrease (p < 0.05), and the highest value (45.70%) was obtained at the P3 level, and the lowest value (10.51%) was obtained at the P1 level. However, as the K application ratio increased, the annual percentage change in alfalfa yield showed a significant decreasing trend (p < 0.05) and reached the maximum value (44.8%) at the K1 level and the minimum value (26.11%) at the K4 level. These results indicated that the appropriate application ratios of N, P, and K of 14.28–27.72%, 36.36–50%, and below 25%, respectively, significantly enhanced the alfalfa yield.

The percentage change in annual alfalfa yield reached its maximum value at the T2 level of NPK application (Table 4 and Figure 11). Moreover, at this level, the application ratios of N, P, and K corresponded to the ratios that produced the maximum annual yields (levels N2, P3, and K1, respectively). Therefore, the optimal NPK application rate was >225–310 kg/ha, and the optimal application ratios of N, P, and K were 14.28–27.72%, 36.36–50%, and below 25%, respectively. These results could be used as the recommendations for fertilization practices to obtain high alfalfa yields in the arid and semi-arid regions of China.

3.5. Contribution of N, P, and K Fertilizer on Alfalfa Yield

In this study, the Aggregated Boosted Tree (ABT) algorithm was used to evaluate the relative importance of N, P, K, and NPK application on alfalfa yield. The analysis results (Figure 12) indicated that the application rates of NPK and K had the greatest (37.61%) and the smallest (15.34%) contribution to annual alfalfa yield, respectively. The importance of the application rates of N and P on the annual alfalfa yields was 22.55% and 24.50%, respectively. By analyzing the effect of fertilization ratio on alfalfa yield, it was found that the P/NPK, N/NPK, and K/NPK rations increased the annual alfalfa yield by 38.64%, 31.71%, and 27.65%, respectively. It could be concluded the cumulative contribution ratios of NPK and P to the annual alfalfa yield exceeded 60%, indicating that these are key factors that affect the annual yield of alfalfa in the arid and semi-arid areas of China.

The factors that influence the first cutting cycle alfalfa yield (in descending order in terms of importance) were NPK (45.86%), N (27.20%), P (14.62%), and K (12.32%). The order of the second cutting cycle was P (43.96%), K (25.36%), N (20.66%), and NPK (10.02%). For the third cutting cycle, the order was NPK (44.92%), N (23.05%), P (19.70%), and K (12.33%). And for the fourth cutting cycle, it was K (42.30%), NPK (28.60%), P (16.93%), and N (12.17%).

4. Discussion

4.1. Effects of N Application on Alfalfa Yield

N is a crucial component in the synthesis of proteins, chlorophyll, and other important substances in alfalfa plants, and it plays a vital role in promoting the growth, yield, and quality of alfalfa [34,35]. Although alfalfa has the ability to fix nitrogen, nitrogen fertilizer is still required to promote nodule formation during the planting stage [36], and studies have shown that the application of N at a rate of 60–120 kg/ha can significantly increase the alfalfa yield in China [15]. In this study, the annual yield of alfalfa increased and then decreased as the N application rate increased. Specifically, the highest values of the mean annual yield and the average annual percentage change in alfalfa yield were 14,162 kg/ha and 22.15% under an N application rate of >50–100 kg/ha, and the lowest values were 10,784 kg/ha and 6.78% under N application rates exceeding 150 kg/ha.

The alfalfa yield of different cutting cycles followed a similar trend and peaked at an N application rate of >50–100 kg/ha. This may be because alfalfa promotes its own growth through a combination of root uptake of nitrogen from the soil and nitrogen fixation using rhizobia under low N conditions (i.e., 0–105 kg/ha), and excessive N application inhibits alfalfa growth [9]. The yield of alfalfa decreased as the cutting cycles increased under N application rates below 100 kg/ha. The yield of different cutting cycles increased and then decreased under N application rates exceeding 100 kg/ha, and peaked at the second cutting cycle, which may be attributed to the excessive N application inhibiting the number of nodules, and the nitrogen-fixing capacity of the nodules [37].

The lower yield at the first cutting cycle may be due to insufficient N application, while the lower yields at the third and fourth cutting cycles may be due to excessive N application. The percentage change in each cutting cycle yield first increased and then decreased, possibly because the symbiotic rhizobia on the roots can fix atmospheric nitrogen and convert it into amino acids that can be absorbed by alfalfa roots to meet their growth needs [38]. The nitrogen-fixing activity of alfalfa roots is mainly observed in the 0–10 cm soil layer, with an annual nitrogen fixation rate of approximately 270 kg/ha [39], and each ton of alfalfa hay harvested removes about 27 kg of soil nitrogen. It is generally believed that alfalfa meets its growth needs through nitrogen fixation by its roots, and the availability of exogenous nitrogen has little influence on its growth process. Although alfalfa can meet part of its nitrogen demand through root nitrogen fixation, the nitrogen-fixing enzyme activity of rhizobia are relatively weak during the early growth and senescence stages, so appropriate N application could significantly promote the growth and yield of alfalfa [40]. Moreover, the nitrogen-fixing enzyme activity of alfalfa roots is higher under suitable N application conditions, which is conducive to nodulation and nitrogen fixation [41] and could also improve winter survival rates. However, excessive N application could reduce the nitrogen-fixing capacity of alfalfa rhizobia, increase the number of ineffective nodules, and inhibit root growth and development, thereby affecting the alfalfa yield [42,43].

4.2. Effects of P Application on Alfalfa Yield

P plays a crucial role in cellular growth, photosynthesis, and other vital processes in alfalfa plants. The application of P provides energy for nitrogen fixation in alfalfa root nodules, but P is easily adsorbed and precipitates in soil, reducing its mobility and leading to a decrease in the available P content in the soil. As a result, the phosphorus in the soil that is available for absorption and utilization by alfalfa roots only accounts for 2–3% of the total soil P content. Soil P stress greatly inhibits the growth and yield of alfalfa. Increased P application could effectively alleviate soil P stress to increase alfalfa yields [44,45].

However, excessive P application can lead to nutrient waste. This study shows that P application can significantly increase alfalfa yield in the arid and semi-arid regions of China, as indicated by the increasing annual yield of alfalfa with increased P application. The highest values for the mean annual yield and the average annual percentage change in alfalfa yield were 14,028 kg/ha and 37.09% under a P application rate exceeding 180 kg/ha, and the percentage change in the first cutting yield was also the highest (39.49%). This conclusion is similar to that of Li et al. [46] in their six-year (2008–2013) field trial in the North China Plain, which found that the highest annual yield of alfalfa was achieved with a P application rate of 210 kg/ha. The results showed that as the P application increased, the yields of the first and second cutting cycles increased, while the yields of the third and fourth cuttings cycles increased and then decreased. Zhang et al. [47] conducted a two-year field trial in Xinjiang, China, and reached a similar conclusion. This may be because the P absorption capacity of alfalfa roots approaches or reaches saturation as the cutting cycle increases, and the P applied to soil is easily bound to certain ions in the soil and becomes fixed, leading to the accumulation of soil P and inhibited of P absorption by alfalfa roots [48].

In addition, the alfalfa yield decreased as the cutting cycle increased under the same P application rate. The alfalfa yield at the first cutting cycle accounted for about 30% of the annual yield, which may be because the roots accumulated sufficient nutrients in the previous year [26] and had undergone a longer growth period under suitable temperature and sunlight conditions. The alfalfa growth of the third cutting cycle roughly occurs in July to August, and excessively high temperatures may inhibit alfalfa yields in the arid and semi-arid regions of China. The fourth cutting cycle had the lowest yield, which may be related to the slow growth of plants due to the shortened daylight hours and the lower temperatures in autumn [49].

4.3. Effects of K Application on Alfalfa Yield

As a key activator of alfalfa photosynthesis, K affects the efficiency of chlorophyll synthesis by regulating enzyme activity and chloroplast structural stability, and K fertilizer deficiency seriously affects alfalfa yield [50,51]. As the amount of K increased, the number of alfalfa root nodules significantly increased, and the enhanced nitrogen-fixing ability subsequently boosted the alfalfa yield [52]. Waldemar et al. [53] found that the K application rate in Poland was 120 kg/ha, and the annual alfalfa production was 14,900 kg/ha. In Wyoming, USA, the annual yield of alfalfa (8632 kg/ha) was 15.25% higher with K application [54]. The results of this study found that the highest values for the mean annual yield and average annual yield percentage change were 13,422 kg/ha and 19.85% under a K application rate exceeding 180 kg/ha. It was also found that the K requirement of high-yield alfalfa was larger.

The yields of the first and second cutting cycles increased as the K application rate increased, while the yield of the third cutting cycle increased and then decreased. This may be because the first and second cutting cycles of alfalfa required more K to facilitate root absorption of N and P [55], and the temperature during the first and second cutting cycles was relatively favorable for growth, increasing the alfalfa’s K demand. However, as the temperature gradually decreased during the third cutting period, excessive K application could negatively impact alfalfa yields.

Additionally, the absorption of K by alfalfa roots gradually decreases as the cutting frequency increases [56]. The yield of alfalfa significantly as the decreased cutting cycle increased under the same K application rate. The analysis of the effects of the K application rate on the percentage change in alfalfa yield of the different cutting cycles revealed that the effect of a K application rate of >120–180 kg/ha on the yield of the first cutting cycle was the strongest (49.18%), while its effect on the third cutting cycle was the weakest and the yield was occasionally lower than that of no K application (−14.45%).

4.4. Effects of NPK Application on Alfalfa Yield

The reasonable application of NPK has a crucial role in producing high alfalfa yields. Yang et al. [57] found that using an NPK amount of 240 kg/ha with an N:P:K ratio of 1:2:1 on alfalfa fields in northern Xinjiang increased the alfalfa yield by 77.31% compared with unfertilized fields. This study showed that an NPK amount > 225–310 kg/ha with application ratios of N, P, and K of 14.28–27.72%, 36.36–50%, and below 25% significant increased alfalfa yields in the arid and semi-arid regions of China, which is in agreement with the findings of Yang et al. [57].

Furthermore, the combined application of NPK was found to have the highest contribution to annual alfalfa yields (37.61%) in this study, which is consistent with the results of Cai et al. [58]. The effect of different nutrient elements on the yield of alfalfa were also different. A possible reason could be the different roles of NPK fertilizers on alfalfa growth; for example, the reasonable application of N promotes leaf photosynthesis and root growth [59], while P application facilitates enhanced photosynthesis and the transport of the photosynthetic products of alfalfa leaves [60] and the combined application of P and K promotes root growth, increases the number of effective root nodules, and enhances nitrogen fixation levels, ultimately increasing alfalfa yields [52].

4.5. Study Limitations and Future Research

The impact of fertilization on alfalfa yield was analyzed in this study, but we did not consider factors such as irrigation, planting duration, variety, and so on. Furthermore, there were differences in the field trial treatments, soil conditions, climate, and other factors across the different studies, which adds uncertainty to the results of this study. Nevertheless, the findings of this research still provide some insights into the effects of N, P, and K fertilizer on alfalfa yield in the arid and semi-arid regions of China, and they can be used as a reference for exploring mechanisms to increase the yield of alfalfa. In the future, more attention should be paid to the effects of fertilizer sources and types, as well as soil characteristics on annual and cutting cycles alfalfa yields.

5. Conclusions

(1)

In the arid and semi-arid regions of China, the annual alfalfa yield increased and then decreased as the N application rate increased, but it increased with increasing P and K application rates. A significant factor affecting the annual and first and third cutting cycles yields was the NPK amount. For the second cutting cycle yield, the effect of P predominates, while K fertilization was the most important factor for the fourth cutting cycle yield.

(2)

The amounts of NPK and P had cumulative contributions to the annual alfalfa yield that exceeded 60%, indicating that these are key factors that influence the alfalfa yield in the arid and semi-arid regions of China. The contribution of P/NPK, N/NPK, and K/NPK to the annual yield of alfalfa decreases successively.

(3)

The optimal fertilization strategy for maximizing alfalfa yields in the arid and semi-arid regions of China was found to be an NPK amount > 225–310 kg/ha with N, P, and K proportions of 14.28–27.72%, 36.36–50%, and below 25%, respectively.