Soil chemical properties affecting soybean yield on a nationwide scale in Japan

ABSTRACT

Wide-area surveys have been conducted on a prefectural or regional scale in Japan for revealing soil chemical properties affecting soybean yield. However, they vary by region and thus, the common chemical factors through Japan remain unclear. This study aims to identify the soil chemical properties involved in soybean yield based on field data in various prefectures throughout Japan. We collected datasets of the topsoil chemical properties and soybean yields at 228 sampling points from 16 prefectures in Japan over the period of 2015–2017. The data sampling was conducted on two fields with different yield levels, namely high- and low-yield groups, which were managed by the same cooperating farmers in each point. We regarded the two fields at each point as a pair for a pairwise comparison. The average yield of the high- and low-yield groups was 3.1 t ha⁻¹ and 2.5 t ha⁻¹, respectively. The p-values of the pairwise comparison results were calculated via the Wilcoxon signed rank test. The differences of the exchangeable Mg (p= 0.040), and the Mg/K (p= 0.027) and Ca/Mg (p= 0.007) ratio between the high-yield and low-yield groups were statistically significant, whereas the other variables were not significant. The exchangeable Mg and Mg/K ratio were negatively related to yield, while the Ca/Mg ratio was positively related. The results suggested that excessive Mg for K or Ca adversely affects the soybean yield. As far as we know, there have been few reports focusing on the relation between the exchangeable cation, especially for the cation balance, and soybean yield. Our results may help researchers and producers deepen their understanding of the relationship between soil chemical properties and soybean yield.

KEY WORDS:

• Wide-area survey
• pairwise comparison
• cation balance
• exchangeable Mg
• Ca/Mg and Mg/K ratio

1. Introduction

The average soybean yield in Japan (1.7 t ha⁻¹) was approximately 60% of worldwide production (2.7 t ha⁻¹) in 2017–2018 (USDA 2019). A total of 80% of the soybean-producing fields in Japan consist of those converted from rice paddies (MAFF 2019); consequently, these fields have physical problems, particularly poor drainage. Accordingly, many studies in Japan have focused on improving the physical properties of the soil to increase soybean yield (Sato et al. 2007; Takahashi 2015; Takahashi et al. 2005, 2018). Additionally, soil chemical properties, such as the indigenous soil N supply and soil pH, are thought to be key factors in increasing the soybean yield in Japan (Hattori et al. 2013; Nakano et al. 1989; Sumida, Kato, and Nishida 2005; Takahashi et al. 2005; Tsubouchi and Saito 2010). However, other chemical properties involved with the yield have not yet been clearly identified.

To investigate the relationships between the soil chemical properties and soybean yield, wide-area surveys have been conducted on a prefectural or regional scale in Japan　(Hattori et al. 2013; Ishitsuka et al. 2016; Odahara et al. 2012; Takahashi et al. 2005, 2014; Yamane and Kokubun 2014). Takahashi et al. (2014) reported that total N in soil was the most significant factor for soybean yield in the Tohoku region in northeast Japan. Takahashi et al. (2005) reported that the amount of mineralized N positively correlates with yield of soybean grown in 33 fields with drainage in the Joetsu area of Niigata Prefecture. Odahara et al. (2012) reported that total N, total C, and the amount of mineralized N positively and pH (H₂O) was negatively correlated with yield in 29 fields in the Chikugo River Basin in Fukuoka. Ishitsuka et al. (2016) reported that exchangeable K saturation was positively correlated with soybean yield by analyzing 214 samples from 10 areas in Fukuoka. These results suggest that the factors involved in soybean yield vary by region. However, wide-area survey on a nationwide scale has not been conducted and thus, factors affecting soybean yield on a larger scale have not been fully understood. This study aims to identify common factors affecting soybean yield throughout Japan based on field data from 16 prefectures.

2. Materials and methods

2.1. Dataset

We collected datasets of the topsoil chemical properties and soybean yields at 228 sampling points from 16 prefectures in Japan over the period of 2015–2017 (Figure 1, Table 1). Since varieties, tillage, cropping, weather conditions, and plant density are considerably different among the sampling sites, simple correlation analysis could not be helpful to reveal nationwide common factors affecting soybean yield and thus, we also employed a pairwise comparison method to cancel the effects of such external factors. Firstly, we selected two fields with different yield levels, namely high- and low-yield groups, at each sampling point; our dataset then consisted of variables from 456 fields. The two fields at each point were managed by cooperating farmers under the same conditions (e.g., varieties, tillage, cropping, weather, plant density). In each field, we investigated the soybean yield (gross seed yield) in an area of around 3 m². The plant density (mean: 15.0 plants m⁻²) in each field ranged from 1.3 to 33.5 plants m⁻² and varied from each farmer. We also collected the topsoil (5.0–30.8 cm, mean: 16.8 cm after molding) at three points and then mixed them. Total N, pH (H₂O), available P₂O₅ (Truog-P), exchangeable K, Ca, and Mg, and the ratios of those cations, i.e., Ca/K, Mg/K, and Ca/Mg were determined (n = 1) by the standard procedures (Kamewada et al. 1997). The collected soils were classified into 10 soil groups based on the work of Obara et al. (2011) (Table 2). The soils were mostly classified as lowland soils (72.0%), followed by andosols (15.3%) and peat soils (6.4%). The dominant varieties of soybean at sampling points were ‘Satonohohoemi,’ ‘Fukuyutaka,’ ‘Enrei,’ ‘Tachinagaha,’ and ‘Nakasennari’ (Table 1). The weather conditions at sampling points from the Agro-Meteorological Grid Square Data, NARO (Ohno et al. 2016) are summarized in Table 3.

Table 1. Number of sampling pairs and varieties in each prefecture over the period of 2015–2017

Prefecture	sampling point				Varieties
	2015	2016	2017	Total
Hokkaido	0	5	5	10	Yukihomare
Aomori	5	5	5	15	Oosuzu, Okushirome
Yamagata	4	5	5	14	Satonohohoemi, Enrei
Miyagi	5	5	5	15	Miyagishirome, Tanrei, Tachinagaha, Suzuhonoka
Niigata	5	5	5	15	Satonohohoemi, Enrei
Ishikawa	6	6	6	18	Satonohohoemi, Enrei
Nagano	6	6	0	12	Nakasennari, Tsubuhomare, Suzuhomare
Ibaraki	3	5	5	13	Tachinagaha, Satonohohoemi
Tochigi	5	5	5	15	Satonohohoemi
Yamanashi	5	5	5	15	Ayakogane, Nakasennari, Akebonodaizu
Aichi	5	8	7	20	Fukuyutaka
Mie	6	10	5	21	Fukuyutaka
Shiga	5	5	5	15	Fukuyutaka, Tamahomare, Ootsuru
Hyogo	3	1	1	5	Sachiyutaka-A1, Tatsumaro, Tanbaguro
Shimane	5	5	4	14	Tamahomare, Sachiyutaka, Fukuyutaka
Fukuoka	0	5	6	11	Fukuyutaka

Table 2. Proportion of soil groups in the sampling sites

Soil group	Proportion (%)
Gley Lowland soils	31.2
Gray Lowland soils	19.8
Lowland Paddy soils	15.4
Brown Lowland soils	3.6
Immature Lowland soils	3.0
Andosols	15.6
Peat soils	5.2
Pseudogley soils	4.4
Brown Forest soils	1.0
Red-Yellow soils	0.8

Table 3. Weather conditions at each growth phase in all fields in 2015–2017

	P1	P2	P3	P4	P5	P6	P7	P8
Daily mean temperature (°C)
Mean	22.8	24.2	25.2	25.8	24.6	21.5	17.3	12.6
Max	29.7	29.9	29.9	29.5	28.4	26.9	24.7	22.0
Min	15.3	16.9	17.5	21.2	20.6	16.2	10.3	4.2
Median	22.3	24.3	25.2	26.0	24.6	21.5	17.2	12.4
SD	3.7	3.1	2.5	1.8	1.8	2.3	2.9	3.4
Daily mean precipitation (mm)
Mean	4.3	4.3	5.0	5.0	6.0	6.1	4.6	3.8
Max	24.0	32.2	30.5	27.5	18.2	18.2	14.2	15.6
Min	0	0	0	0	0.7	0.5	0.4	0.1
Median	3.0	3.4	3.2	4.0	5.1	5.3	4.2	3.3
SD	4.2	4.2	5.2	4.3	3.9	3.6	2.8	2.4
Daily mean daylight hour (hr)
Mean	5.7	5.9	5.9	6.1	5.4	4.7	4.6	4.8
Max	10.3	10.6	11.6	11.5	8.6	8.3	7.8	7.8
Min	1.1	1.9	0.9	1.1	1.1	1.6	1.4	2.1
Median	5.6	5.7	5.6	6.2	5.5	4.7	4.4	4.8
SD	1.9	1.9	2.2	2.1	1.7	1.5	1.3	1.2
Daily mean solar radiation (MJ m^−2)
Mean	18.2	18.4	18.1	18.1	16.0	13.4	11.3	9.6
Max	24.5	24.7	25.2	25.2	21.1	20.6	15.8	16.0
Min	10.3	11.8	9.7	8.2	8.6	8.3	6.9	4.8
Median	18.1	18.3	18.1	18.4	16.3	13.7	11.2	9.4
SD	3.0	2.8	3.1	3.2	2.9	2.4	1.9	2.0

All data was acquired from the Agro-Meteorological Grid Square Data, NARO (Ohno et al. 2016) (https://amu.rd.naro.go.jp/).

P1–P4: days from sowing to flowering is equally divided into four phases.

P5–P8: days from flowering to harvest is equally divided into four phases.

Figure 1. 228 sampling points from 16 prefectures involved in the wide-area survey

2.2. Statistical analysis

We firstly calculated Pearson’s correlation coefficients (r) between each chemical property of soils and the yield regardless of yield level (n = 438–456) using R (Ver 3.5.1). We also compared with each pair of the variables in our dataset and calculated the p-value via the Wilcoxon signed rank test (Sokal and Rohlf 1994; Wilcoxon 1945) using the wilcox.exact function in the exactRankTests package (Hothorn and Hornik 2019) in R (Ver 3.5.1).

3. Results

3.1. Yield and chemical properties of soils

Table 4 shows the mean, maximum, minimum, and median values of each variable in the high- and low-yield groups. The mean yield of the high-yield groups was 0.6 t ha⁻¹ higher than that of the low-yield groups. However, differences in the mean total N, pH (H₂O), available P₂O₅, and exchangeable K and Ca were less than 2% between the two groups. The mean Ca/K and Ca/Mg ratios of the high-yield groups were 2.7% and 7.5% higher than those of the low-yield groups, respectively, whereas the mean exchangeable Mg and Mg/K ratio of the high-yield groups was 7.1% and 6.0% lower than that of the low-yield groups, respectively.

Table 4. Descriptive statistics for relatively high- and low-yield groups in 2015–2017

Variable	Units	Mean	Max	Min	Median	SD	n	Normality
High-yield fields
Yield	t ha^−1	3.14	5.45	1.18	3.17	0.75	228	0.21
Total N	g kg^−1	2.24	7.64	0.83	1.90	1.17	219	<0.05
pH (H2O)	dimentionless	5.84	7.50	4.59	5.80	0.55	228	<0.05
Available P2O5	mg kg^−1	218	1182	2	181	152	225	<0.05
Exchangeable K	mg kg^−1	275	1046	43	232	180	220	<0.05
	cmolc kg^−1	0.70	2.68	0.11	0.60	0.46	220	<0.05
Exchangeable Ca	mg kg^−1	1822	5000	130	1647	881	220	<0.05
	cmolc kg^−1	9.09	24.95	0.65	8.22	4.40	220	<0.05
Exchangeable Mg	mg kg^−1	267	784	45	213	159	220	<0.05
	cmolc kg^−1	2.19	6.45	0.37	1.75	1.31	220	<0.05
Ca/K	dimentionless	16.8	71.3	0.9	13.7	11.0	220	<0.05
Mg/K	dimentionless	4.03	20.85	0.33	3.24	3.26	220	<0.05
Ca/Mg	dimentionless	4.98	19.08	1.50	4.41	2.64	220	<0.05
Low-yield fields
Yield	t ha^−1	2.50	4.97	0.20	2.51	0.87	228	<0.05
Total N	g kg^−1	2.21	7.50	0.82	1.81	1.16	219	<0.05
pH (H2O)	dimentionless	5.85	7.50	4.50	5.84	0.58	228	0.10
Available P2O5	mg kg^−1	217	2260	8	167	202	225	<0.05
Exchangeable K	mg kg^−1	273	890	41	232	167	220	<0.05
	cmolc kg^−1	0.70	2.28	0.10	0.59	0.43	220	<0.05
Exchangeable Ca	mg kg^−1	1824	4931	245	1692	907	220	<0.05
	cmolc kg^−1	9.10	24.61	1.22	8.44	4.53	220	<0.05
Exchangeable Mg	mg kg^−1	287	927	16	249	174	220	<0.05
	cmolc kg^−1	2.36	7.63	0.13	2.05	1.43	220	<0.05
Ca/K	dimentionless	16.4	71.2	1.3	13.9	10.3	220	<0.05
Mg/K	dimentionless	4.29	21.84	0.39	3.44	3.49	220	<0.05
Ca/Mg	dimentionless	4.64	18.96	1.25	3.99	2.46	220	<0.05

Normality: p-values calculated by the Shapiro-Wilk test.

3.2. Pearson’s correlation coefficient and Wilcoxon signed rank test

We calculated Pearson’s correlation coefficients (r) between each chemical properties of soil and the yields regardless of yield level (n = 438–456; Figure 2, Table 5). Although the exchangeable K (r = −0.11) and Ca/K ratio (r = 0.15) were significant, the r values were very low and thus the correlations were considered negligible as we expected.

Table 5. Pearson’s correlation coefficient between yield and each variable on data of both yield fields

Variable	Correlation coefficient (r)	p-value	Significance
Total N	0.07	0.16
pH (H2O)	−0.02	0.64
Available P2O5	−0.02	0.65
Exchangeable K	−0.11	0.02	*
Exchangeable Ca	0.07	0.16
Exchangeable Mg	−0.03	0.56
Ca/K	0.15	0.001	**
Mg/K	0.06	0.24
Ca/Mg	0.09	0.07

*: p < 0.05, **: p < 0.01, ***: p < 0.001

Figure 2. Relation between each chemical property of soil and the yield regardless of yield level

Table 6 shows the result of calculating the p-value of each variable using the Wilcoxon signed rank test. The differences of the exchangeable Mg (p= 0.040), and the Mg/K (p= 0.027) and Ca/Mg (p= 0.007) ratio between the high-yield and low-yield groups were statistically significant, whereas the other variables were not significant.

Table 6. Wilcoxon signed rank test for each variable in pairs of high- and low-yield groups

Variable	Difference	Paired test
		p-value	Significance
Yield	0.66	< 0.001	***
Total N	0.09	0.172
pH (H2O)	−0.04	0.230
Available P2O5	14	0.182
Exchangeable K^a	0.01	0.644
Exchangeable Ca^a	−0.22	0.825
Exchangeable Mg^a	−0.30	0.040	*
Ca/K	−0.20	0.927
Mg/K	−0.20	0.027	*
Ca/Mg	0.42	0.007	**

*: p < 0.05, **: p < 0.01, ***: p < 0.001

Difference = (median in high-yield groups) – (median in low-yield groups)

a: cmol_c kg⁻¹

4. Discussion

The very low r (−0.11–0.15) between each property of soils and yields (Table 5) might result from ignoring the significant differences in the varieties, tillage, cropping, and weather condition among sampling points (Tables 1–3). The previous wide-area surveys (Ishitsuka et al. 2016; Odahara et al. 2012; Adams et al. 2017, 2018; Villamil, Davis, and Nafziger 2012) also have ignored such external factors that can be strongly associated with the yield of crops, causing a misinterpretation of the relationship. Our results by the pairwise test might be more robust than the previous wide-area studies (Ishitsuka et al. 2016; Odahara et al. 2012; Adams et al. 2017, 2018; Villamil, Davis, and Nafziger 2012), which may contribute to the improvement of soybean yields.

There have been no consistent results for the effect of Mg on the yield of soybean to this time. Mg is bound to the center of chlorophyll involving in plant photosynthesis at leaves, and thus some studies have used the Mg fertilizer to improve soybean yields (Reinbott and Blevins 1995; Nelson, Burkhart, and Colwell 1946; Altarugio et al. 2017). Our results, however, showed that the mean exchangeable Mg in the high-yield groups was significantly lower than that in the low-yield groups (Table 6). Some previous reports also showed the negative effect of Mg on the soybean yield (Adams et al. 2017; Cox et al. 2003; Ishitsuka et al. 2016; Villamil, Davis, and Nafziger 2012). Cox et al. (2003) analyzed the relationship between soil properties and soybean yields and found a significant negative correlation between the exchangeable Mg and the yield in two of three fields. Villamil, Davis, and Nafziger (2012) and Adams et al. (2017) also showed a significant negative effect of Mg on soybean yields in U.S.A. In addition, some previous reports showed that the exchangeable Mg had no significant effect on soybean yields (Adams et al. 2018; Ishitsuka et al. 2016; Odahara et al. 2012; Sawchik and Mallarino 2008; Cox et al. 2003). Only a few studies have shown a positive correlation of the exchangeable Mg with the yield, whereas this relationship has not always been found (Sawchik and Mallarino 2008; Jiang and Thelen 2004). The previous and present studies suggest that the exchangeable Mg may be a negative factor for the soybean yield in Japan.

The Mg/K (p= 0.027) and Ca/Mg (p= 0.008) ratios were significant factors for the soybean yield in our datasets (Table 6). The average of the former in the high-yield groups was significantly lower than that in the low-yield groups, whereas the latter was completely the opposite (Table 6). In addition to this, p-value was in the order of exchangeable Mg (p= 0.040) > Mg/K (p= 0.027) > Ca/Mg (p= 0.007) (Table 6), suggesting a possibility that excessive Mg for K or Ca adversely affects the soybean yield. Little has been reported on the relationship between the exchangeable cation ratio and the yield (Takahashi et al. 2014). Takahashi et al. (2014) reported that the Ca/Mg ratio was positively correlated with the yield at the 10.4% significance level. However, the exchangeable Mg also had a positive correlation with the yield at the same significance level, contradicting the results of the Ca/Mg ratio (Takahashi et al. 2014). In addition, the balance of exchangeable Ca, Mg, and K generally has only a small influence on the crop yield (Kopittke and Menzies 2007). Nevertheless, the concentration of Ca in the soybean leaves at the flowering stage has been positively correlated with soil Ca/Mg, seemingly contributing to the yield (Tsubouchi and Saito 2010). Thus, further studies are required to determine how much the cation balance influences soybean yield.

Soil pH is closely related to the exchangeable cations as well as the chemical composition of the soil and could contribute to the soybean growth. In Japan, the recommended pH (H₂O) range is from 6.0 to 6.5 for upland crops (MAFF 2015). Tsubouchi and Saito (2010) reported that pH (H₂O) was positively correlated with N₂ fixation activity and water-soluble molybdenum content in soil, which contributed to the yield. However, although the soil pH (H₂O) values were lower than 6.0 in half of the studied fields (Table 4), the pH (H₂O) was not different significantly between the high-yield and low-yield groups (Table 5), suggesting that the pH (H₂O) was not related to the soybean yield. Some wide-area studies in Japan showed that pH (H₂O) is not significantly correlated with soybean yield (Ishitsuka et al. 2016; Takahashi et al. 2014; Yamane and Kokubun 2014). Similarly, Adams et al. (2018) conducting the wide-area survey in Arkansas concluded that the pH (H₂O) values were not significantly different between high-yield (4.7–6.4 t ha⁻¹) and average-yield (3.8–5.6 t ha⁻¹) areas. In addition, some reports showed the negative effect of soil pH (H₂O) on the soybean yield (Odahara et al. 2012; Villamil, Davis, and Nafziger 2012). The previous and present wide-area studies suggest the negative or no effect of soil pH on the soybean yield in Japan.

Acknowledgments

We are grateful to members of the consortium entitled “Development of diagnostic methods and countermeasure techniques for overcoming high yield inhibitory factors.” and local farmers in each prefecture for collecting the detailed data in the present study. We also thank several anonymous reviewers for providing feedbacks on our paper.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This study was supported by the Ministry of Agriculture, Forestry, and Fisheries of Japan under a grant entitled ‘Development of diagnostic methods and countermeasure techniques for overcoming high yield inhibitory factors.’