Yield and dry matter production of soybean response to late planting in southwestern Japan

ABSTRACT The selection of the sowing date is one of the most important decisions in soybean production. It is important for farmers to know the expected seed yield when developing a sowing plan. The objective of this study was to quantitatively reveal the decreased seed yield ratio during late sowing in southwestern Japan under irrigated conditions. Total above ground dry matter and distribution were measured from July to August sowing in 2018 and 2019 in Fukuyama, Hiroshima, Japan. The harvest index in the August sowing did not decrease compared to the July sowing under irrigated conditions. The decrease in yield was related to the total aboveground dry matter rather than to the harvest index. The decrease in total aboveground dry matter was related to the total amount of solar radiation intercepted rather than to radiation use efficiency. A significant regression equation was obtained for the relative yield, which was calculated from the maximum seed yield observed from early to mid-July. When seeds were sown after mid-July, our result showed that the seed yield will decrease by 0.60% per day for Sachiyutaka A1 and by 0.64% for Akimaro, even under conditions of 100% seedling establishment and controlled soil water conditions. This regression equation can be used in fields as an indicator of seed yield where irrigation is possible. GRAPHICAL ABSTRACT


Introduction
Soybean is an important food crop in Japan; therefore, ensuring a high and stable seed yield is essential.Selection of the sowing date, planting densities, and cultivars are among the most important decisions in soybean production.It is important for farmers to know the expected seed yield when developing a sowing plan.As the cultivation area of a single farmer has increased recently, a judgment criterion is required to redevelop the sowing plan when the sowing date is expected to be delayed.In particular, there have been many reports of unstable sowing dates for soybeans in southwestern Japan, such as due to torrential rains in 2018 and a prolonged rainy season in 2020.
Soybeans generally bloom under short-day conditions and several analyses have been conducted to predict the seed yield under different sowing dates in the U. S (De Bruin & Pedersen, 2008;Kessler et al., 2020;Salmerón et al., 2015).In Japan, late sowing shortens the entire growth duration and decreases seed yield (Kawasaki et al., 2018;Kumagai & Takahashi, 2020;Takeda & Sasaki, 2013).
However, several reports have shown that the difference in seed yield between normal and late sowing is not statistically significant (Ikejiri & Takahashi, 2016;Isobe et al., 2020;Matsuo et al., 2013;Suzuki et al., 2017).Ikejiri and Takahashi (2016) reported that the difference in seed yield was not significant because of the larger seed size in the July sowing than in the June sowing in Yamaguchi prefecture.An increase in seed number per pod during late sowing has been observed in the south Kanto region (Isobe et al., 2020;Suzuki et al., 2017).These studies indicate that the effects of late sowing are complicated and there is still room to discuss the effects of late sowing on dry matter production and distribution.
The authors compared June sowing and July sowing in Fukuyama under irrigated conditions and reported that the total aboveground dry matter at R8 is related to the total amount of solar radiation intercepted and that dense planting is effective in maintaining seed yield in July sowing (Kawasaki et al., 2018).In addition, the harvest index increased during July sowing under irrigated conditions in southwestern Japan.However, information on August sowings is limited in Japan.Takeda and Sasaki (2013) pointed out the possibility of poor seedling establishment from late July to mid-August sowing in southwestern Japan due to severe drought.Although irrigation is a solution for establishing seedlings, it often requires human resources and time.Therefore, it is important to determine the maximum seed yield during August sowing under well-irrigated condition.Kumagai and Takahashi (2020) reported that the decrease in seed yield in late sowing was related to a lower harvest index in northeastern Japan.In southwestern Japan, the harvest index may decrease due to lower temperature or lower solar radiation in the late seed-filling stage in extremely late sowing, such as August sowing.
Recently, a process-based model was developed to predict soybean development (Nakano et al., 2020(Nakano et al., , 2021)).Although regression analysis is a basic approach, as shown by Takeda and Sasaki (2013), a more detailed analysis is required for seed yield prediction adopted in wide areas.In the fields of local farmers, many factors that cause a decrease in seed yield have been identified during late sowing, such as poor seedling establishment, shorter vegetative growth duration, lower solar radiation in the seed-filling stage, drought, and wet injury.In soybean, drought stress and wet injury inhibit dry matter production by decreasing photosynthetic rate, stomatal conductance, nodule nitrogen fixation or canopy coverage (Bajgain et al., 2015;Shimada et al., 2012).Appropriate control of water table can reduce soil moisture related stresses (Shimada et al., 1995(Shimada et al., , 2012)).In addition, poor seedling can be solved by underground irrigation (Takeda & Sasaki, 2013).Therefore, the maximum seed yield will be strongly influenced by shorter vegetative growth duration associated with short daylength, and lower solar radiation in the seed filling stage under irrigated condition in late sowing.It is important to determine the meteorological potential seed yield of domestic cultivars under well-irrigated conditions as an indicator of farmers' sowing plans.
The objective of this study was to quantitatively reveal the decreased ratio of seed yield in late sowing under uniform seedling establishment and controlled soil moisture to support farmers' development of cultivation plans in southwestern Japan.To achieve this objective, dry matter production and distribution data were collected from a wide range of sowing dates.

Growth condition
Field experiments were conducted in 2018 and 2019 at Western Region Agricultural Research Center, NARO, Fukuyama, Hiroshima, Japan (34°30′N, 133°23′ E).Two cultivars (Sachiyutaka A1 gou and Akimaro) were grown in a lysimeter (alluvial soil).The singleplot area of the lysimeter was 3.7 × 3.7 m (11 rows × 23 plants).Seeds were sown on July 2, 13, 24, and August 6 in 2018, and on July 5, 19, August 2, and 16 in 2019.Two seeds were seeded in each well.Seeds were sown in rows by hand at 0.3 m between rows with 0.15 m intra-row spacing (22.2 plants m −2 ).Two replicates were analyzed each year.After seedling emergence, the plants were thinned to one plant per hole and the gaps were filled by transplanting.Inorganic fertilizers were applied as a basal dressing at 3 g m −2 of N, 6 g m −2 of P 2 O 5 , and 6 g m −2 of K 2 O.In addition, fertilizer of micronutrients, containing 19% of manganese (Mn) and 9% of boron (B) (FTE, TOMATEC CO., LTD) was applied 6 g m −2 before sowing in 2018 because the value of easily-reducible Mn in soil was low in 2017 (Table 1).After sowing, the water table was set 10 cm below the soil surface to ensure a uniform emergence date.After emergence, the water table was maintained 35 cm below the soil surface.Instead of intertillage, hand weeding was conducted until canopy closure.Insecticides and fungicides were applied periodically to avoid biotic stress.

Measurement
Daily solar radiation and temperature data were recorded at Western Region Agricultural Research Center, NARO, Fukuyama.The growth stage (R1, R5, R7, R8; Fehr & Caviness, 1977) were recorded.The fraction of solar radiation was estimated using canopy coverage according to Purcell (2000).Canopy coverage was measured by digital image analysis using ImageJ software ver.1.50 once or twice a week after emergence until canopy closure.Cumulative intercepted solar radiation was estimated by interpolating the daily canopy coverage and cumulating daily intercepted solar radiation according to Shiraiwa et al. (2011).Solar radiation use efficiency (RUE) was determined by dividing total above ground dry matter at R8 by the cumulative intercepted solar radiation from emergence to R7. Seed yield and total above ground dry matter at R8 were determined from 40 plants (1.8 m 2 ) harvested from one replicate.To calculate moisture content, stems and pod shells were weighed before and after 72 h oven-drying at 80°C.Seed yield was expressed at 15% moisture content.The harvest index was calculated from the seed dry matter and total aboveground dry matter, excluding the leaves and petioles attached at R8.The yield components (total number of nodes per plant, total number of pods per plant, number of fertile seeds per pod, and 100-seed weight) were estimated from six representative plants in each replicate.In addition to the yield components, the main stem length, number of branches, and number of main stem nodes were measured.The crude protein concentration in the grains was measured with a near infrared spectrometer (Infratec 1241 Grain Analyser, FOSS JAPAN Ltd., Tokyo, Japan) using a protein conversion factor of 6.25.
Quantifying the rate of decrease in seed yield during late sowing was an important objective of this study.Therefore, standardization was attempted to integrate the two years of data.Salmerón et al. (2015) calculated 'relative yield,' defined as the ratio to the maximum seed yield of year and location for each cultivar, to discuss data from multiple locations.In this study, we defined 'relative yield' as the ratio of the maximum seed yield of year for each cultivar.All statistical analyses were conducted using Bell Curve for Excel software ver.2.15 (Social Survey Research Information Co. Ltd., Tokyo, Japan).

Results
Table 2 shows the changes in the mean air temperature, solar radiation, and cumulative precipitation at Western Region Agricultural Research Center, NARO in 2018 and 2019.Torrential rain was observed in early July 2018 in southwestern Japan, including Fukuyama.After the torrential rains, the cumulative precipitation was low until late August 2018.The mean temperature in the early growing season, from early July to late August, was 1.8°C higher in 2018 than in 2019.However, over the whole growing season, from early July to late November, the mean temperature was 0.1°C higher in 2018 than in 2019.In 2018, the solar radiation from early July to late August was 4.5 MJ m −2 d −1 higher than in 2019 but the solar radiation from early September to late October in 2019 was 1.9 MJ m −2 d −1 higher than in 2018.
Table 3 shows the observed emergence and reproductive growth stages for the sowing dates in 2018 and 2019.Emergence (VE) was observed on July 6, 17, and 28 and August 10 in 2018, and July 9 and 22 and August 6 and 19 in 2019.Both Sachiyutaka A1 gou and Akimaro reached maturity (R8) normally at all sowing dates in the experiment conducted in Fukuyama.No snow or frost was observed during the entire growth period in this experiment.Late sowing shortened the growth period (VE to R7).Shortening of the growth period occurred not only from germination to R1 but also at all stages from R1 to R5 and R5 to R7. Akimaro reached each growth stage later than Sachiyutaka A1 gou on all sowing dates but there was no clear difference between the cultivars in the shortening of the growth period due to late sowing.There were large differences between two years in seed yields.Although coefficients in 2019 were not statistically significant, both cultivars tended to produce lower seed yields with late sowing (Figure 1).As for harvest index, some coefficients were not statistically significant.But the harvest index tended to increase with late sowing (Figure 2).Although the harvest index tended to be high after late sowing in 2019, the difference in seed yield in the same year was related to the difference in total aboveground dry matter rather than the difference in harvest index, and the total aboveground dry matter at R8 was lower after late sowing (Figure 3).In the same year, the difference in total aboveground dry matter at R8 among the sowing dates was related to the cumulative intercepted solar radiation rather than the RUE (Figure 3).In other words, the decrease in cumulative intercepted solar radiation by late sowing reduced the total aboveground dry matter and seed yield.
Table 4 shows yield components Sachiyutaka A1 gou and Akimaro at each sowing date.Main stem length, number of mainstem nodes, number of total pods, and 100-seed weight decreased in August sowing in both cultivars.On the other hand, pod number per node and crude protein did not show clear trends.The seed number per pod increased in late July and early August.
The results of the single regression analysis of emergence date (day of the year) and 'relative yield' and the results of the single regression analysis of cumulative intercepted solar radiation and relative yield were statistically significant (P < 0.05) (Figures 4 and 5).Fehr and Caviness (1977).

Discussion
In this experiment, growth stages were collected from a wide range of sowing dates.The differences in the R1-R5 period between the first sowing and the second sowing in Akimaro are larger than that of Sachiyutaka A1 gou, which suggests the differences in the combination of maturing genes.However, both Sachiyutaka A1 gou and Akimaro reached R8 on all sowing dates (Table 3).Takeda and Sasaki (2013) did not identify the date of maturity in August sowing at the experiments conducted in local farmers' fields in Okayama prefecture.In our experiment, although a decrease in seed yield was observed in late sowing (Figure 1), both early and mid-August sowing reached maturity in Fukuyama, Hiroshima prefecture.Kumagai and Takahashi (2020) reported that a decrease in the harvest index was observed during late sowing in northeastern Japan, in addition to a decrease in total aboveground dry matter.Chilling injury in the early reproductive stages, such as flowering or pod setting, is a serious problem that severely decreases seed yield (Kurosaki & Yumoto, 2003).On the other hand, in our experiment in southwestern Japan, the harvest index did not clearly decrease in the August sowing (Figure 2).Kurosaki and Yumoto (2003) observed a severe decrease  in pod setting under low-temperature treatment, during which the temperature settled to 18°C (day-time temperature) and 13°C (night-time temperature) in the early reproductive stage.In Fukuyama, Sachiyutaka A1 gou and Akimaro in August sowing lasted to R5 in September and the mean air temperature of September was over 20°C (Table 2).It is supposed that early reproductive development, such as flowering or pod setting, was not a limiting factor in seed yield compared with dry matter production in our experiment.
Low temperatures cause chilling injury, delayed maturity, or poor grain filling in rice, even in warm regions (Sato, 1967).While chilling injury, such as a decrease in pods per node, was not observed in this study, 100-seed weight in the August sowing decreased (Table 4).Even if the August sowing matures normally, careful decisions are required because a small seed size can decrease prices.In addition to poor grain filling, areas with high humidity due to snowfall may make it difficult to harvest soybean by machines.In northern Japan, the first frost days are focused on avoiding frost damage to soybeans (Sameshima et al., 2007).Although no snow or frost was observed during the entire growth period in our experiment, the first frost day can be an indicator in southwestern Japan, especially in mountainous or northern areas.
In this study, we clarified that the decrease in seed yield due to late sowing was caused by a decrease in dry matter production rather than a decrease in the harvest index from early July to mid-August under irrigated conditions (Figures 2 and 3).The difference in total aboveground dry matter at R8 between the two years shown in Figure 3 was related to the difference in RUE rather than the cumulative intercepted solar radiation.The difference in the RUE between the two years was thought to be influenced by the soil environment, particularly by the amount of easily-reducible Mn.The soil sampled after the soybean harvest in 2017 was deficient in Mn and B; therefore, fertilizer of micronutrients (FTE) was applied before sowing in 2018 (Table 1).Fertilization was conducted only in 2018 and not before the soybean sowing in 2019, but the results of the soil chemical analysis showed an increasing trend not only in the fall of 2018 but also until the fall of 2019.In particular, the increase in easily-reducible Mn was larger in 2019 than 2018.Because FTE is soluble in citric acid, it is conceivable that the effect of the FTE applied in 2018 May have been delayed and showed its effect in 2019.Because Mn is involved in the biosynthesis of chlorophyll, it can be hypothesized that a deficit in Mn decreased the RUE in 2018.However, the negative effect of the Mn deficit was included in the effect of year, such as differences in meteorological conditions, and could not be detected in this study.The maintenance of soil fertility is an important issue in soybean production.While a relative comparison was conducted in this study, it is important to associate dry matter productivity and soil fertility in the future to precisely predict seed yield.
In this study, we attempted standardization by analyzing the relationship between relative yield and sowing date.The results of the regression analysis were statistically significant (P < 0.05) (Figure 4).When seeds were sown after mid-July, our result showed that the seed yield will decrease by 0.60% per day for Sachiyutaka A1 and by 0.64% for Akimaro, even under conditions of 100% seedling establishment and controlled soil water conditions (Figure 4).With 15% moisture content. 2) Values followed by the same letters were not significantly different based on Tukey's multiple comparison at the 5% level.
Under irrigated conditions, the variation in the harvest index among the sowing dates was smaller than that in the total aboveground dry matter.Our results indicated that the degree of seed yield reduction can be estimated from the cumulative intercepted solar radiation under irrigation conditions (Figure 5).These results are expected to contribute to farmers' sowing plans as a potential seed yield in a field that can be sufficiently irrigated.Although we measured canopy coverage in this study, if combined with a model that predicts leaf area and vegetation cover development using meteorological data (Nakano et al., 2020), it would be useful to predict the potential seed yield precisely in the case of late sowing at any site.
In fields where irrigation is not possible, poor seedling establishment during late sowing is considered a serious problem.Takeda and Sasaki (2013) reported that very few seedlings were established in fields where irrigation was not possible in early August.If irrigation is not possible, it is a matter of concern that decrease in seedling establishment, photosynthesis rate (Chomsong et al., 2020), RUE, and harvest index (Kawasaki et al., 2013) due to water stress may occur.Therefore, the precise quantification of seed yield reduction under drought stress is required in future studies.
An important issue that could not be examined in this study was earlier sowing.Early sowing is expected to result in longer vegetative growth periods and higher dry matter production.In this study, late sowing caused little variation in the harvest index but it is conceivable that the harvest index may decrease because of an increase in the proportion of vegetative growth organs, such as larger stem dry matter (Kawasaki et al., 2018), and smaller pod setting ratio (Kawasaki et al., 2018).Although severe lodging was not observed in this study, it can be a serious problem in early sowing as the main stem length increases.

Conclusions
A quantitative analysis was conducted for the delayed sowing of soybean in southwestern Japan.The harvest index of sowing in mid-August did not decrease compared with that in early July.The decrease in seed yield was related to the total aboveground dry matter, rather than to the harvest index.The decrease in the total aboveground dry matter was related to the total amount of solar radiation intercepted rather than radiation use efficiency.A significant regression equation was obtained for the relative yield, which was calculated from the maximum seed yield observed from early to mid-July.This regression equation can be used as an indicator of seed yield for fields in which irrigation is possible.On the other hand, the decreased ratio in fields where irrigation is impossible can be larger because of poor seedling establishment and drought stress.If combined with a water stress index or drought indicator, the decrease in seed yield in farmers' fields can be precisely estimated.The range of our experiment is not sufficient to reveal the effect of frost and snow, which sometimes decrease seed yield or make machine harvesting difficult and may occur in mountainous areas or northern areas in southwestern Japan.Although the results of our  regression analysis can be used to develop a cropping plan, further studies are needed to evaluate the effects of chilling injury.

Figure 1 .
Figure 1.Relationships between emergence date (day of the year) and seed yield.(a) relationship of sachiyutaka A1gou and (b) relationship of Akimaro.Error bars show standard error (n = 2).r means Pearson's correlation coefficient.*, † means significant at P < 0.05, P < 0.10, respectively.NS means non-significant at P = 0.10 level.

Figure 2 .
Figure 2. Relationships between emergence date (day of the year) and harvest index.(a) relationship of sachiyutaka A1gou and (b) relationship of Akimaro.Error bars show standard error (n = 2).r means Pearson's correlation coefficient.† means significant at P < 0.10.NS means non-significant at P = 0.10 level.

Figure 4 .
Figure 4. Relationship between emergence date (day of the year) and relative yield.

Figure 5 .
Figure 5. Relationship between cumulative intercepted solar radiation (MJ) and relative yield.

Table 1 .
Soil chemical properties of the experimental site (alluvial soil) from 2017 to 2019.

Table 2 .
Changes in mean air temperature, solar radiation and precipitation at Fukuyama in 2018 and 2019.

Table 3 .
Growth stages of Sachiyutaka A1 gou and Akimaro at each sowing environment.

Table 4 .
Seed yield and yield components of Sachiyutaka A1 gou and Akimaro at each sowing environment.