Effects of Tillage System, Sowing Date, and Weather Course on Yield of Double-Crop Soybeans Cultivated in Drained Paddy Fields

Abstract

In temperate monsoon areas, major constraints of soybean production in drained paddy fields are excess soil water during monsoon seasons. To further understand how agronomic practices and weather course affect the yield of soybeans, we conducted field experiments at Gwangju, Korea (35°10′ N, 126°53′ E) over three years (2018–2020). Double-crop soybeans were grown at two tillage systems (TS) [rotary tillage (RT), deep plowing followed by rotary tillage (DPRT)] and three sowing dates (SD) (June 10–15, June 25–30, and July 10–15) in drained paddy fields. Flowering phenology (R2) was accelerated by 5 days with each 15-day delay in SD. This resulted in a significant reduction in vegetative growth up to R2, with subsequent reductions in CGR and NAR through R2–R5 (beginning of grain filling). With a 30-day delay in SD, yield was significantly reduced by 27.0%. The better performance of RT over DPRT was demonstrated by greater yields (13.7%). In addition, yield was greatly varied with weather volatility among years, ranging from 123.8 to 552.0 g m⁻². Weather volatility was the greatest contributor to yield variability (30.4%), followed by SD (17.0%) and TS (10.3%). Our results suggest that the yield might be mainly determined by how much growth has already been achieved before flowering and through R2–R5.

1. Introduction

Over the last dozens of years, rice productions have been outpacing consumptions, which are still steadily declining in some East Asian countries such as Korea and Japan. Due to such changes in rice production/consumption and changes in government policies for food production, farmers are switching their key paddy crops from rice in flooded conditions to upland crops (e.g., soybean, maize, and other economic crops) in drained paddy conditions. Among upland crops, soybean (Glycine max (L.) Merr.) is the most widely grown (30–80%) one in drained paddy fields across East Asian countries [1] due to its nutrient and economic values [2]. However, excess soil water frequently limits soybean productions in drained paddy fields because these fields are usually geographically positioned in low-lying areas [3]. In addition, with the ongoing global warming, which is projected to be further intensified due to the global hydraulic cycle [4], the stability of soybean productions in poorly drained paddy fields is becoming increasingly vulnerable to excess soil water including flooding and soil waterlogging [5,6,7]. In the United States, the world’s biggest soybean producer, soil waterlogging can cause a soybean yield loss of 9–93% depending on waterlogging durations and crop stages [3]. In addition, soybean yield losses by excessive soil water can differ depending on a number of agricultural conditions, such as management practices [8], topography and soil types [3], tillage [9,10], crop rotation [9], sowing date and density [11,12], and cultivars [13].

In Korea located in a temperate monsoon zone, double-crop soybeans, which are inclined to have late sowing (hence having more chance to be influenced by monsoon during vegetative to early reproductive stages), could suffer more from excessive soil water, especially in poorly drained paddy fields. The monsoon not only can lead to excess soil water but also can lead to a lack of sunshine. In addition, climate warming expected in the near future (~2050) is projected to enhance heavy rainfall events by 24.0% and inter-annual variability by 31.5% in Korea [14]. Despite such concerns, less attention has been paid to the inter-annual yield variability of double-crop soybean cultivated in drained paddy fields exposed to weather volatility over the years. Most studies on temporal yield variability of soybean have been conducted in uplands for either monoculture or rotation crops [8,9,15]. Soybean yields in the United States showed an inter-annual variability of 20–48% [8,9,15] in distinctly different growing conditions. Such previous results indicate that annual variability of soybean yield could be affected by multi-factors such as years (i.e., annual weather volatility), regions, and agricultural management, which make it difficult to pinpoint limiting-factor effects and thus the ability to seek ideal growing conditions [16].

A compacted soil layer, which is generally developed by no-tillage [17] and wheel traffic of agricultural machinery [18] can reduce porosity and hydraulic conductivity [17,19]. Such soil conditions can reduce water infiltration and soil volume explored by roots, thus restricting root development in deeper soil layers [19,20]. Moreover, no-tillage and/or compacted soil could persist in flooding conditions for a long time when fields were flooded, unless otherwise counteracted [19]. On the contrary, tillage disrupts the compacted layer and creates favorable soil environments so that roots can develop well in the deeper soil layer. It provides a greater opportunity for roots to absorb water and nutrients with a greater root exploration, possibly leading to increased yield [19,20,21,22]. Although whether no-tillage has advantages for improving crop productivity [10,17,19,20,21,22,23,24] and sustainable land use [25] remains controversial, soil tillage can reduce soil bulk density and enhance soil volume explored by root, thus improving root development. However, fewer studies have reported the impacts of tillage systems (TS) on soil water dynamics, growth, and yield of upland crops such as soybean cultivated in drained paddy fields [10]. There has been a growing need for more informative knowledge about what TS will have a better performance for stabilizing and improving the productivity of double-crop soybean in poorly drained paddy fields.

In the case of mono-cropping soybean, sowing dates (SD) can appropriately be adjusted either to avoid abiotic/biotic stresses or to make full use of available weather resources for increasing crop growth and yields [3]. However, the SD of double-crop soybean in temperate monsoon areas is usually delayed due to wet soils following commencing monsoon season just after harvesting prior crops. Consequently, it will compel soybeans to grow under conditions with a short day-length and a high temperature during the early stages at least. Because soybean is a short-day plant, crop phenology responds sensitively to photoperiod. It also responds well to higher temperatures. Actually, several studies have reported that soybean crops can advance their phenology under conditions of short-day [26,27], warming [25,28,29,30], and interactions between these two factors [26,27]. Recent evidence for soybean, rice, and maize has also suggested that each 1 °C increase in temperature can lead to a shortening of the vegetative period by more than 3 days [25]. Another recent study has reported that late SD by 20 days shortens the period from emergence to the beginning of flowering by 7 days, with an increase in mean temperature of 2.48 °C and a decrease in cumulative day-length of 95.7 h [26]. As such, later SD is most likely to cause a shorter crop duration and thus a lower yield [3]. However, additional studies are essential to better understand how SD affects crop phenology, growth, and yield processes of double-crop soybeans cultivated in drained paddy fields.

The present study addresses the following key questions that have not been fully answered yet in relation to growth and yield responses of double-crop soybeans grown in drained paddy fields to TS and SD coupled with weather volatility: (1) How does SD alter crop phenology and subsequent growth and yield processes of soybeans in drained paddy fields? (2) To what extent does the yield of soybeans vary with weather volatility between crop seasons? (3) How do TS and SD affect the growth and yield of soybeans? To answer these questions, we conducted field experiments for double-crop soybean as a sequential crop of winter barley under drained paddy fields over three crop seasons.

2. Materials and Methods

2.1. Study Cite Descriptions

The study site was located at drained paddy experiment fields of Chonnam National University, Gwangju, Korea (35°10′ N, 126°53′ E). Background soil (0–15 cm) properties of the drained paddy are summarized in

Table 1. Background soil properties of experimental drained paddy fields used for this study.

Parameters	Values
pH (1:5)	6.83
Total C (g kg−1)	11.0
Total N (g kg−1)	1.2
Available P (mg P2O5 kg−1)	5.6
CEC (cmol kg−1)	7.2
Texture	Silty clay loam
Sand (g kg−1)	105
Silt (g kg−1)	608
Clay (g kg−1)	287

pH was measured at a 1:5 ratio of soil to water; total C and N were measured with a combustion method by Nelson and Sommers [31]; available P was determined with the Bray #1 method by Kuo [32]; CEC was determined with NH₄OAc method by Sumner and Miller [33]; texture was determined based on the USDA classification after particle-size fractionation using the pipette method by Gee and Bauder [34].

. The soil was classified as coarse loamy, mixed, nonacid, mesic family of Fluvaquetic Endoquepts in the Soil Taxonomy. The area, together with the neighboring Jeollanam-do province, has a typical agro-climate that grows a large proportion of the Korean rice crop, occupying >20% of the total area harvested (approximately 8 × 10⁵ ha). This site is subjected to the East Asian temperate monsoon climate system with an annual mean temperature of 14.1 °C over the past 30 years (1991–2020, hereafter referred to as the “normal year”). Annual mean precipitation was about 1380.6 mm, raining about 55.6% of the time from June to August (monsoon season) when double cropping soybeans are approximately in the stages from sowing to early reproductive phases.

2.2. Field Establishments and Treatments

Experiments were conducted at drained paddy fields (20 m wide and 60 m long). These fields were established for studying the double cropping of upland crops (i.e., soybean and winter barley). Before starting the experiments, we established main drainage channels (40 cm width and 40 cm depth) around all drained paddy fields using a compact excavator to promote drainage without a stagnation of excess water between ridged rows for soybean, hence avoiding or minimizing water stresses for plants during the monsoon season. In this study, we planted soybeans as a double-crop following winter for barely over three years (2018–2020). There were two levels of TS and three levels of SD. Thus, the experiments consisted of six treatment combinations. The experimental design had a split-plot arrangement of a randomized complete block with three replications. TS was whole plot treatments and SD was sub-plot treatments. Plots of 600 m² (10 m × 60 m) and 400 m² (20 m × 20 m) were assigned every year for each of the TS and SD treatments. The TS consisted of (1) rotary tillage (RT) to a depth of 12 cm, and (2) deep plowing to a depth of 25 cm followed by rotary tillage (DPRT). Chisel tillage with four blades was used for deep plowing. RT and DPRT were performed according to the schedule of SD treatments after harvesting barely in early June every year. Target SD treatments were performed on June 10 (the earliest SD possible), 25 June and 10 July 2018 and 15 June, 30 June, and 15 July 2019 and 2020. Due to soil conditions before and after rainfall events, the actual SD of some target SD (i.e., 10 July 2018 and 15 July 2019 and 2020) were either advanced or delayed by 1–3 days from target SD treatments. Nonetheless, across three years, actual SD treatments were performed on 10–15 June, 25–30 June, and 10–15 July (hereafter referred to as “S1”, “S2”, and “S3”, respectively). Considering soil conditions and weather forecast, tillage and ridge constructions were made on the day of sowing or otherwise a few days before sowing. Prior to tillage, all treatment plots were fertilized with 3.0 g N-3.0 g P₂O₅-3.4 g K₂O m⁻² based on results of soil tests just after barely harvesting every year. For both RT and DPRT, ridges with a width of 110 cm and a height of 20 cm were constructed by making a furrow (30 cm width and 20 cm depth) after rotary and leveling (Figure 1). This was done to raise the root system out of the saturated layer and minimize plant water stresses during monsoon seasons, thus creating a favorable soil environment for root growth. All other agronomic practices including weed and pest control were similar to those used by local farmers.

2.3. Crop Managements and Measurements

In all TS (RT and DPRT) and SD (S1, S2, and S3) plots, 3 to 4 soybean seeds (cv. Daewon with determinate growth habit) were sown on a ridge in two rows with a 70 cm row space and 10 cm hill space (equivalent to 14.28 hills m⁻²; Figure 1). Two plants per hill were established by thinning after emergence. To know soil water dynamics in RT and DPRT plots, multi-sensor subsurface soil probes (EnviroPro^®, APCOS Pty Ltd., SA, Australia) were installed between rows in the S1 plot of both RT and the DPRT in 2018. Soil water profiles in volumetric water content (%) in four soil layers at intervals of 10 cm from the surface were monitored season-long every 10 min and stored in the data logger (CR1000, Campbell Scientific Inc., Logan, UT, USA) as an average value of one hour. After emergence, key development stages (i.e., crop phenology) for all treatment plots were carefully monitored throughout the whole season based on the Fehr and Caviness scale [35]. At R2 (flowering) and R5 (beginning of grain filling), leaf area index (LAI, m² m⁻²) was non-destructively measured with a plant canopy analyzer (LAI-2200C, LI-COR Inc., Lincoln, NE, USA) in 2018 and 2019. In 2020, however, it was destructively measured with an automation area meter (AAM-9, Hayashidenko Co. Ltd., Tokyo, Japan) due to extremely long-term rainfall events (see Section 2.4), which rarely allowed us to employ a plant canopy analyzer. In all three years, to determine crop growth and yield parameters, areas of the soybean crop were destructively sampled at R2, R5, and R8. For all treatment plots, 15 hills were sampled from above the ground at three replication blocks. Plants were separated into green leaves (when present, leaf area was measured in 2020), main stem (including branches and petioles), and pods (when present). Biomasses of plant parts were determined separately after drying at 80 °C in a forced air oven to a constant weight (>72 h). Specific leaf areas (SLA, cm² g⁻¹) at R2 and R5 were calculated by dividing total leaf area by total leaf biomass. The biomass was presented as aboveground biomass (AGBM, g m⁻²) by converting it into grams per square meter based on planting density. The final biomass at R8 was presented as apparent AGBM after petioles and leaves had fallen. Based on changes in AGBM (dw) and LAI (dLAI) at two samplings (dt; R5 and R2) and mean LAI (LAI_m; Equation (3)), CGR (crop growth rate, Equation (1)) and NAR (net assimilation rate, Equation (2)) were calculated as follows:

CGR = dw/dt

(1)

NAR = CGR/LAI_m

(2)

LAI_m = dLAI/(lnLAI_R5 − lnLAI_R2)

(3)

For all plants sampled at R8, the number of nodes on the main stem was counted, and yield and its parameters were investigated. Fine grain excluding immature and damaged seeds was weighed. Yields converted into unit of land area are presented as fine grain weight at a 13% moisture.

2.4. Weather during Growing Seasons

In normal years (1991–2020), mean temperature, rainfall, and sunshine durations during double-crop soybean seasons corresponding to the current study were 22.0–22.8 °C, 647–931 mm, and 5.5–5.9 h, respectively (

Table 2. Weather conditions during the whole season and each growth period with different sowing dates of double-crop soybeans grown in drained paddy fields over three years (2018–2019) and normal year (1991–2020).

Year	Crop Stage	Mean Temperature (°C)			Rainfall (mm)			Rainfall Events (day)			Sunshine Duration (h d−1)
		S1	S2	S3	S1	S2	S3	S1	S2	S3	S1	S2	S3
2018	WS	23.4	23.3	22.4	1437	1415	1130	42	39	31	6.8	6.7	7.3
2019	WS	23.5	23.2	23.1	732	623	572	44	40	36	5.6	5.5	5.6
2020	WS	22.6	22.9	22.9	1559	1456	1206	51	45	37	5.0	4.8	5.2
Normal (1991–2020)	WS	22.8	22.7	22.0	931	840	647	-	-	-	5.5	5.6	5.9
2018	S–R2	26.7	28.4	29.5	307	304	90	11	9	4	7.1	7.3	9.0
	R2–R5	28.9	26.6	23.7	90	470	885	4	13	11	8.1	6.1	4.8
	R5–R8	19.0	17.4	16.4	1040	641	155	27	17	16	5.9	6.4	6.7
2019	S–R2	24.7	26.3	27.4	352	261	222	17	16	14	5.4	5.2	5.7
	R2–R5	28.3	26.0	24.1	30	48	68	5	5	9	7.0	5.4	2.9
	R5–R8	21.3	20.3	19.8	350	314	282	22	19	13	5.2	5.7	6.1
2020	S–R2	23.7	24.4	26.2	637	1206	956	26	28	21	3.5	2.5	4.0
	R2–R5	27.5	28.1	25.0	661	65	178	7	3	8	5.5	7.2	4.4
	R5–R8	20.2	19.8	18.8	261	185	72	18	14	8	6.1	6.0	6.6
Normal (1991–2020)	S–R2	24.9	26.0	26.8	438	448	371	-	-	-	4.9	4.9	5.5
	R2–R5	26.8	25.5	24.1	210	185	129	-	-	-	5.5	5.2	5.3
	R5–R8	19.9	18.6	17.5	284	208	147	-	-	-	6.1	6.3	6.4

S1, S2, and S3 indicate sowing dates of 10–15 June, 25–30 June, and 10–15 July, respectively. WS, whole season (i.e., sowing to R8) at different sowing dates; S, sowing.

). Weather conditions during growing seasons in the current study (2018–2020) showed an obvious volatility among three years: a close to normal in 2018, less rainfall and sunshine duration in 2019, and wet due to heavy rainfall with a high frequency of precipitation events (thus less sunshine duration) in 2020 (Table 2). Of 23 days from 20 July to 12 August, in particular, it rained for 20 days including raining for eight consecutive days twice in 2020. In addition, mean temperature, rainfall and its event days, and sunshine durations at each growth stage showed a clear distinction between years as well as SD (Table 2). In the present study, all weather and climate data used observations from Gwangju Station (Station ID 156, 35°10′ N, 126°53′ E) of the Korean Meteorology Administration, near (<650 m) our experimental fields.

2.5. Statistical Analysis

For statistical analysis, the experimental design was treated as a blocked split-split plot. Experimental year was the whole-plot treatment. Levels of TS (RT and DPRT) had a split-plot treatment and levels of SD (S1, S2, and S3) had a split-split plot treatment. A linear mixed model (Equation (4)) was used to estimate the effects of year, TS, SD, and their interaction on growth and yield parameters of double-crop soybean grown in drained paddy fields:

Z_ijkl = µ + B_i + Y_j + (BY)_ij + T_k + (BT)_ik + (YT)_jk + (BYT)_ijk + S_l + (BS)_il + (YS)_ji + (BYS)_ijl + (TS)_kl + (BTS)_ikl + (YTS)_jkl + (BYTS)_ijkl + e_ijkl

(4)

where Z_ijkl is the response variable at block i (=0, 1, 2), year j (=0, 1, 2), TS_k, (=0, 1), SD_l (=0, 1, 2) within a subplot; µ is the overall mean; B, Y, T, and S are the block, year, TS, and SD effects, respectively; and e_ijkl is the error term. The model was fitted using the restricted maximum likelihood (REML) procedure in SAS Mixed (v. 9.4, SAS Institute, Inc., Cary, NC, USA) in order to estimate the means and standard errors for each combination of TS and SD. Treatment means for measured parameters were compared using Fisher’s least significant difference (LSD, α = 0.05). The coefficient of variation (CV), which could be calculated by dividing average yield by the standard deviation of yield, was used to assess temporal crop yield variability for each treatment plot. Soybean yields are also presented with CV as boxplots, in which lower and upper hinges represent the first and third quartiles (the 25th and 75th percentiles). Interquartile ranges (space between first and third quartiles) were plotted with both median and mean values.

3. Results

3.1. Soil Water Profiles in Different Tillage Systems and Soil Depths

Soil water profiles differed largely between TS over both soil depth and season (Figure 2a,b). Soil layer based on volumetric soil water content, except early stage, was divided approximately into two layers of 10 cm and 20–40 cm in RT (Figure 2a), while in DPRT it was divided into three layers of 10 cm, 20 cm, and 30–40 cm (Figure 2b). Soil water content over the season varied from 5.2% to 32.4% at 10 cm and from 29.3% to 70.5% at 20–40 cm in RT. It varied from 4.8% to 49.3% at 10 cm, 17.7% to 76.8% at 20 cm, and 36.1 to 71.2% at 30–40 cm in DPRT. In addition, the soil water content in each depth pulsed more sensitively in DPRT as soil water was altered by either rainfall, irrigation, or evapotranspiration compared to those in RT. The maximum water content was commonly observed at 20 cm in RT and at 30 cm DPRT, except when it temporarily shifted to 40 cm in RT and 20 cm in DPRT with heavy rainfall (>40 mm h⁻¹) events occurring in the second half of the season (Figure 2c). On average, soil water contents ranged from 22.0% at 10 cm to 57.6% at 30 cm in DPRT and from 16.7% at 10 cm to 44.8% at 20 cm in RT, exhibiting a greater soil water content in DPRT than in RT over the season.

3.2. Crop Phenology and Growth Parameters at Different Tillage Systems and Sowing Dates

Days to flowering (R2) were significantly hastened as SD was delayed across three years (Figure 3d–f). However, they did not differ between TS. With every 15-day delay in SD, soybean reached R2 by 5.5, 5.0, and 5.0 days earlier in 2018, 2019, and 2020, respectively, exhibiting a greater response of flowering phenology to SD than to cropping years (Figure 3d–f). This was a primary reflection of both increased growth temperature and decreased day-length up to R2 as SD delayed (Figure 3a–c). Mean growth temperatures up to R2 increased by 1.1, 1.6, and 0.7 °C in 2018, 2019, and 2020, respectively, when the SD was delayed from S1 to S2. When SD was delayed from S2 to S3, mean growth temperatures up to R2 increased by 1.1, 1.1, and 1.8 °C in 2018, 2019, and 2020, respectively. In contrast, the day-length shortened by 0.18–0.19 h and 0.26–0.34 h across the three years when SD was delayed from S1 to S2 and from S2 to S3, respectively (Figure 3a–c).

Up to R2, the production of aboveground biomass (AGBM) mirrored the response of R2 to SD, except that S1 and S2 in 2018 showed a slight delay in emergence due to herbicide (Figure 3g). A similar response of AGBM to SD was also observed at R5 and R8 (

Table 3. Growth parameters of double-crop soybeans grown in drained paddy fields with different tillage systems and sowing dates over three seasons (2018–2020). ANOVA (analysis of variance) results (p-values) are given for each parameter.

Year	Tillage	SD	LAI (m2 m−2)		SLA (cm2 g−1)		NAR (g m−2 d−1)	CGR (g m−2 d−1)	AGBM (g m−2)
			R2	R5	R2	R5	R2–R5	R2–R5	R5	R8
2018	RT	S1	2.4 bc	4.1 b	187.7 ab	134.3 c	9.17 a	29.59 a	902.1 a	911.2 a
		S2	2.1 c	4.5 b	153.9 b	181.0 bc	6.00 abc	18.90 abc	774.3 a	832.6 ab
		S3	3.6 a	4.9 ab	252.0 a	252.2 ab	3.30 bc	13.74 bc	572.4 b	648.1 b
	DPRT	S1	2.8 b	4.8 ab	207.8 a	160.3 c	7.45 ab	27.54 ab	901.7 a	732.8 ab
		S2	2.3 bc	4.6 ab	209.5 a	200.0 bc	5.56 abc	18.07 abc	706.0 a	765.2 ab
		S3	3.7 a	5.4 a	258.3 a	327.4 a	1.96 c	8.85 c	498.1 b	585.2 b
2019	RT	S1	6.1 a	6.3 b	348.2 b	328.1 a	3.76 b	23.29 ab	602.5 a	840.8 a
		S2	6.4 a	7.1 a	512.7 a	323.2 a	2.84 b	19.12 b	639.8 a	697.7 b
		S3	4.7 b	6.1 b	500.7 a	375.8 a	2.61 b	13.96 b	431.0 b	537.7 c
	DPRT	S1	6.0 a	6.2 b	460.8 a	309.5 a	5.98 a	36.26 a	627.7 a	747.2 ab
		S2	6.2 a	7.2 a	504.4 a	383.9 a	2.37 b	15.71 b	559.8 a	684.4 b
		S3	4.5 b	5.5 c	508.6 a	355.2 a	2.75 b	13.73 b	408.3 b	486.4 c
2020	RT	S1	5.4 a	6.3 a	390.0 b	377.0 b	1.93 a	11.24 a	556.6 a	628.5 a
		S2	5.2 ab	6.4 a	632.9 a	441.0 ab	2.49 a	14.23 a	419.2 ab	645.3 ab
		S3	4.4 b	5.6 a	647.7 a	606.3 a	1.65 a	8.18 a	261.6 b	408.3 bc
	DPRT	S1	5.0 ab	6.2 a	511.2 a	386.6 b	2.54 a	14.14 a	523.6 a	347.0 bc
		S2	4.8 ab	6.3 a	660.6 a	516.1 a	2.26 a	12.72 a	394.0 ab	497.2 ab
		S3	4.2 b	5.7 a	569.6 a	603.5 a	1.58 a	7.76 a	267.7 b	335.4 c
ANOVA results
Year (Y)			0.0003	<0.0001	0.0004	0.0003	0.0009	0.0079	0.0021	0.0045
Tillage (T)			0.5363	0.4821	0.3038	0.2004	0.7505	0.9010	0.4224	0.0041
Sowing (S)			0.0062	0.0118	0.0004	<0.0001	0.0001	0.0002	<0.0001	<0.0001
Y × T			0.0947	0.0474	0.9757	0.7457	0.3022	0.6063	0.9369	0.3598
Y × S			<0.0001	0.0002	0.0318	0.0415	0.0063	0.1667	0.5486	0.1724
T × S			0.8782	0.8408	0.1170	0.5552	0.6982	0.3439	0.8316	0.0345
Y × T × S			0.9709	0.6642	0.4838	0.7002	0.6597	0.6898	0.9791	0.7279

SD, sowing date (S1, S2, and S3 indicate sowing dates of 10–15 June, 25–30 June, and 10–15 July, respectively); LAI, leaf area index; SLA, specific leaf area; NAR, net assimilation rate; CGR, crop growth rate; AGBM, aboveground biomass; RT, rotary tillage; DPRT, deep plowing followed by rotary tillage. Different letters within a column in each year indicate statistically significant differences (p < 0.05) across tillage systems and sowing dates.

). Across three years and TS, AGBM ranged from 524 to 902, 394 to 774, and 261 to 572 g m⁻² in S1, S2, and S3 at R5, respectively. They ranged from 347 to 911, 497 to 832, and 335 to 648 g m⁻² in S1, S2, and S3 at R8, respectively. LAI significantly varied with cropping year, SD, and their interactions, ranging from 2.1 to 6.4 at R2 and 4.1 to 7.2 at R5 across treatments (Table 3). For the reason mentioned above, except in 2018, as SD was delayed from S2 to S3 (but not from S1 to S2), LAI decreased on average by 14.1% (2020) to 26.9% (2019) at R2 and by 11.2% (2020) to 18.5% (2019) at R5. On the contrary, SLA significantly increased as SD was delayed. This effect varied with cropping years. With every 15-day delay in SD, SLA increased by about 23.3, 12.5, and 18.8% on average at R2 and by about 40.7, 7.1, and 25.9% at R5 in 2018, 2019, and 2020, respectively (Table 3). However, TS had no significant effect on LAI, SLA, or other growth parameters. During the period from R2 to R5, NAR exhibited a significant decline with a delay in SD. Regarding the extent to which the NAR was decreased with a delay in SD, on average, NAR was decreased by 42.5%, 21.8%, and 12.9% for every 15-day delay of SD in 2018, 2019, and 2020, respectively. It showed a significant difference among cropping years. Responses of CGR to SD and cropping year almost echoed those of NAR, resulting in 37.1%, 31.0%, and 17.4% reduction on average for every 15-day delay in SD in 2018, 2019, and 2020, respectively.

3.3. Yield and Related Parameters at Different Tillage Systems and Sowing Dates

Across TS and SD treatments, yields ranged from 373 to 552 (mean 457, median 473), 286 to 492 (383, 402), and 124 to 329 (242, 240) g m⁻² in 2018, 2019 and 2020, respectively, with variability (CV) of 30.4% (Figure 4a). For TS, yields ranged from 199 to 552 (387, 477) and 124 to 479 (334, 373) g m⁻² in RT and DPRT, respectively, showing a variability of 10.3% (Figure 4b). For SD, they ranged from 124 to 552 (397, 442), 280 to 495 (395, 402), and 221 to 377 (290, 298) g m⁻² in S1, S2, and S3, respectively, with a variability of 17.0% (Figure 4c).

Yields averaged across all treatments varied from 124 to 552 g m⁻² (

Table 4. Yield and its key components of double-crop soybeans grown in drained paddy fields under different tillage systems and sowing dates over three seasons (2018–2020). ANOVA (analysis of variance) results (p-values) are given for each parameter.

Year	Tillage	SD	Pod No. (hill−1)	Seed No. (pod−1)	Fine Seed No. (hill−1)	100 Seeds wt. (g)	Yield (g m−2)
2018	RT	S1	86.4 a	1.82 a	153.6 a	25.3 b	552.0 a
		S2	78.4 ab	1.74 bc	130.1 b	26.6 b	494.7 ab
		S3	62.1 c	1.65 c	98.2 c	26.9 ab	377.0 b
	DPRT	S1	75.7 ab	1.77 ab	130.5 ab	25.5 b	479.4 a
		S2	66.4 c	1.68 c	106.7 c	28.5 a	465.9 a
		S3	62.3 c	1.67 c	97.2 c	26.7 b	373.4 a
2019	RT	S1	89.6 a	1.76 a	158.5 a	23.0 bc	492.4 a
		S2	73.0 b	1.72 a	119.7 b	25.6 ab	404.3 b
		S3	55.0 c	1.73 a	92.9 cd	25.0 abc	308.6 c
	DPRT	S1	79.7 ab	1.72 a	135.2 ab	22.1 c	405.1 b
		S2	70.7 b	1.70 a	114.9 bc	26.2 a	400.3 b
		S3	49.2 c	1.72 a	82.0 d	25.9 ab	286.3 c
2020	RT	S1	59.3 a	1.98 a	110.0 a	21.3 ab	329.3 a
		S2	54.3 ab	1.94 a	99.2 ab	22.8 a	323.8 a
		S3	41.6 bcd	1.97 a	72.5 bc	19.1 c	198.8 b
	DPRT	S1	30.4 d	1.91 a	44.7 d	20.0 b	123.8 c
		S2	48.6 abc	1.94 a	87.0 abc	22.5 a	280.2 a
		S3	39.7 cd	1.97 a	70.3 cd	19.3 bc	193.2 bc
ANOVA results
Year (Y)			0.0054	0.0001	0.0073	0.0005	0.0018
Tillage (T)			0.0011	0.2999	0.0001	0.6588	0.0014
Sowing (S)			<0.0001	0.1928	<0.0001	<0.0001	<0.0001
Y × T			0.2710	0.9911	0.0646	0.4045	0.1308
Y × S			0.0001	0.2208	0.0003	0.0034	0.0035
T × S			0.0066	0.65598	0.0029	0.2070	0.0009
Y × T × S			0.0828	0.9571	0.0736	0.6648	0.2317

SD, sowing date (S1, S2, and S3 indicate sowing dates of 10–15 June, 25–30 June, and 10–15 July, respectively); RT, rotary tillage; DPRT, deep plowing followed by rotary tillage. Different letters within a column in each year indicate statistically significant differences (p < 0.05) across tillage systems and sowing dates.

). Overall, the greatest and lowest yield was observed in 2018 and 2020, respectively (Figure 4a and Table 4). This was primarily due to different weather conditions among the years (Table 2). On average, the yield was 13.7% greater in RT than in DPRT (Figure 4b and Table 4). It exhibited a 27.0% decrease with a delay in SD from S2 to S3, but not from S1 to S2 (Figure 4c; Table 4). Yield reductions due to delay of SD were greater in DPRT than in RT (Table 4). In 2020, DPRT resulted in the lowest yield in S1 (i.e., earliest SD) due to flooding caused by torrential rain (>250 mm d⁻¹) at R2. Yield reductions caused by late sowing also differed among years (Table 4). Pod number per hill decreased significantly with DPRT and with a delay in SD. Seed number per hill also decreased, although seed number per pod remained unchanged (Table 4). The greatest seed number per pod was observed in 2020 when the lowest pod number per hill was observed. Overall, responses of such yield parameters to each treatment alone or treatment interactions exhibited a similar pattern to those of yield (Table 4). Hundred-seed weight remained unchanged with TS, whereas it was significantly increased with a delay in SD from S1 to S2 (but not from S2 to S3).

The number of nodes on the main stem was lower in 2020 than in 2018 and 2019, ranging from 12.9 to 13.5 (mean 13.3, median 13.2), 12.1 to 13.1 (12.5, 12.4), and 10.2 to 13.2 (11.7, 11.6) in 2018, 2019 and 2020, respectively (Figure 5a). Nodes did not differ with TS (Figure 5b), while it declined with a delay in SD from S2 to S3, but not from S1 to S2 (Figure 5c).

Overall, pod number per hill was closely correlated with CGR from R2 to R5 (Figure 6a). A similar correlation was also observed for yield (Figure 6b).

4. Discussion

Our results from three-year experiments for double-crop soybean cultivated in drained paddy fields revealed a huge annual variability of yield. This was primarily attributable to weather volatility among years. We also found that SD had a great potential to cause a big yield variability of soybean cultivated as sequential double cropping (i.e., barely-soybean) in drained paddy fields. In addition, a substantial part of the variation in yield was due to TS possibly affecting soil water distribution and nutrient dynamics [3]. Overall, both weather volatility between cropping years and SD explained a large part of crop phenology, growth, and yield variabilities, with TS having a less contribution to yield variability than cropping years and SD.

4.1. Sowing Date in Tandem with Weather, Rather than Tillage Systems, Alters Flowering Phenology and Growth

We observed that earlier flowering events were associated with a delay in SD, showing 5 days earlier flowering for every 15-day delay in SD within the range from 10–15 June (earliest SD, S1) to 10–15 July (latest SD, S3) in drained paddy field (Figure 3d–f). However, we could not find evidence that TS affected flowering phenology. Crop phenology of soybean, a representative short-day crop, responds sensitively to short day-length and high temperature [25,28,29,30]. In the study site (35°10′ N 126°53′ E), with each 15-day delay in SD, mean growth temperature and day-length (averaged over three cropping seasons) from sowing to flowering showed an increase of 1.2 °C and a decrease of 0.25 h, respectively (Figure 3a–c). Given that soybean is a typically sensitive crop to short-day and warm climates, our results suggest that an increase in growth temperature in tandem with a shortened day-length can accelerate flowering phenology when SD is delayed. A number of reports have also pointed out that soybean crops can advance their phenology with warming [25,28,29,30], short-day [26,27], and interactions between these two factors [26,27]. Recent evidence for soybean, rice, and maize has suggested that each 1 °C increase in temperature can shorten the vegetative period by more than three days [25]. Another recent study has reported that delaying the SD by 20 days can shorten the period from emergence to the beginning of flowering by 7 days, with a 2.48 °C rise in mean temperature and with a reduction in cumulative day-length by 95.7 h [26]. Overall, our results were consistent with previous findings [25,26,27], suggesting that late sowing of double-cropping soybean in drained paddy fields could universally lead to an earlier flowering phenology by increasing growth temperature and shortening day-length at the early stages of the crop season.

The majority of AGBM accumulation at flowering can be considered a result of vegetative growth. Therefore, reductions in AGBM at R2 with an SD delay in 2019–2020 found in this study were most likely the result of a shortage of vegetative growth (Figure 3h,i). The extent to which the AGBM was reduced with a delay in SD was also amplified (e.g., from 29.2 to 45.9%) as SD was delayed from 15 to 30 days. In this study, double-crop soybeans with a delay in SD mostly experienced a higher growth temperature than optimal temperatures of 20–25 °C [36,37] up to at least R2. It has been frequently reported that a temperature higher than optimal is associated with decreases in soybean photosynthetic rate, AGBM, and yield [30,36,38,39]. Similar results have also been documented, showing that a 1–3 °C increase in temperature could cause reductions in AGBM by 11–27% [38]. In some cases, only a 0.4 °C increase in air temperature and a 0.7 °C increase in soil temperature could lead to a decrease in leaf photosynthetic rate (e.g., 6.6–10.3%) at the flowering and seed filling stages [30]. Given the observed increase in growth temperature (1.2 °C per 15-day delay in SD) and the decrease in day-length (0.25 h per 15-day delay in SD) with an SD delay, the majority of reduced AGBM at R2 might be due to shortened vegetative durations. This in turn may directly or indirectly lead to a decline in subsequent AGBM accumulations at R5 and R8 (Table 3). With an SD delay, such serial decreases of AGBM accumulations might be associated with hastened phenology-induced reductions in LAI and reduced LAI-related decrease in CGR during the period from R2 to R5 (Table 3). For example, the shortened vegetative period (i.e., hastened flowering phenology) with an SD delay might impose a decrease on LAI observed at R2 (2019–2020), although the negative effect appeared to be somewhat alleviated at R5. On the other hand, considering noticeable increases in SLA observed in an extremely wet year of 2020, excess soil water might increase the SLA of soybean and lead to a reduction in nodule nitrogen fixation, causing reduced leaf nitrogen and NAR [40].

4.2. Tillage Systems and Sowing Date Coupled with Weather Volatility Lead to Yield Variability

Our results revealed that the yield of double-crop soybean grown in drained paddy fields was different depending on TS and SD as well as weather volatility. Yields averaged across TS and SD ranged from 123.8 to 552.0 g m⁻² over three years (Figure 4a–c and Table 4). The major part of the variation in yield was due to weather volatility between cropping years. Indeed, there was a distinct annual volatility in weather events characterized by a wet but high sunshine duration in 2018, a more or less dry but close to normal sunshine duration in 2019, and an extreme wet coupled with a number of precipitation events (hence less sunshine durations) in 2020 (Table 2). Furthermore, in 2020, torrential rain (>250 mm d⁻¹) observed on 7–8 August (i.e., day corresponding to R2 for the S1) caused temporarily flooding, probably contributing to the lowest yield in 2020 (Figure 4a and Table 4). Overall, such an annual weather volatility well accounted for the annual variability of 30.4% in yield (Figure 4a). This study also showed that different SD and TS caused 17.0 and 10.3% variabilities of soybean yield, respectively (Figure 4b,c). Hence, in terms of the extent to which factors caused yield variability in the present study, annual weather volatility showed the highest extent, followed by SD and the TS. Of several weather events, precipitation events causing excessive soil water have often been reported to reduce yield and lead to spatiotemporal variability of soybean grown in drained paddy fields and/or at low-lying topographic positions with poorly drained soils [3,6,7,8,11]. However, yield reduction and variability due to excessive soil water have shown to be largely varied depending on when and how long crops are exposed to excessive soil water [3]. They also depend on crop management practices [3,8]. For example, Kaur et al. [3] have documented that soybean yield losses due to waterlogging (i.e., soil conditions that soil pores are saturated with water) can vary from 9% to 93% depending on waterlogging durations of 2–14 days, suggesting a greater yield loss in waterlogging at reproductive stage than at vegetative stage [3,7]. Our results showed that soybean yields averaged across TS and SD were significantly lower by 36.8–47.1% in an extremely wet year (2020) than in the 2018–2019 crop seasons. These were within the range reported by previous studies [3,7].

Unlike what was expected from a minor influence of TS on vegetative growth parameters (Table 3), the better performance of RT over DPRT was demonstrated by a greater yield of 13.7% (Figure 4b). The reason for this is currently unclear. DPRT might have led to fast water infiltration, resulting in faster subsoil water recharge. Therefore, it might have caused soybean to be somewhat susceptible to soil waterlogging by holding high levels of volumetric soil water contents as depicted in Figure 2. Such soil water conditions can probably cause reductions in both pod number and seed number per plant [7], resulting in a lower yield in DPRT than in RT. Recently, Ploschuk et al. [7] have reported a significant reduction (−57%) in pod number and yield of soybeans exposed to soil waterlogging conditions between R1 and R3. Similarly, other studies have demonstrated that reductions in pod number and soybean yield are most likely to be due to excessive soil water, including soil waterlogging and flooding [41,42]. DPRT might have also resulted in a reduction in soil N availability that could be declined with soil waterlogging through runoff, leaching, denitrification [3], and inhibiting nodulation [7,13,43]. This could probably be another evidence of greater yield reduction observed in DPRT than in RT. Our results suggest that, even for soybeans grown on a ridge, DPRT for drained paddy fields has the potential to increase the risk of yield reductions due to excessive soil water in extremely wet years, although the extent of the risk can be varied depending on SD and crop stage. Nevertheless, given the soil water profile shown in Figure 2, DPRT may have great potential to increase soil water availability by promoting water storage through soil layers, especially when surface irrigation and water savings are needed. With regard to TS effects on soybean yield, few consistent points of view have been found so far. This might be because its effects are readily varied due to complicated interactions of investigated sites and/or topographies [9,44], management practices [3,19,22,23], crop stages [10], cropping systems [9,45], and weather events [10,23,43,45]. Therefore, further efforts are needed to better understand the pros and cons of multiple TS coupled with a wide range of abiotic and biotic factors for stabilizing the production of double-crop soybeans in drained paddy fields.

With SD delay, rising growth temperature during the vegetative period together with photoperiod change (Figure 3a–c) might have acted not only as a growth constraint (Figure 3h,i and Table 3) but also as a yield constraint (Figure 4c and Table 4) for our double-crop soybean. As summarized in Figure 7, rising growth temperature and shortened photoperiod with SD delay reduced vegetative growth period through hastening flowering phenology, constraining subsequent overall soybean growth, development of reproductive organs, and yield. Late-planted soybean in high rainfall areas such as the Gulf Coast Region (30° N) shows losses of CGR and yield because of waterlogging stress in sensitive periods [11]. In this study, however, we could not find any evidence that late-sowing induced reductions in growth and yield due to greater rainfall or extreme events compared to earlier sowing (Table 2). Hence, our results suggested that reduced vegetative growth (i.e., LAI and AGBM) at R2 with SD delay was most likely to be responsible for subsequent reductions in growth and yield parameters throughout crop stages since flowering (Figure 3h,i, Table 3, and Figure 7). Among weather events, which varied with crop stages, markedly less sunshine durations during the period from R2–R5 were recorded with SD delay (Table 2). This coupled with reduced LAI and AGBM at R2 might have contributed to the reductions in NAR and CGR during the period from R2–R5, leading to subsequent growth and yield reductions. As for CGR, in particular, this study showed that the CGR from R2 to R5 had strong correlations with pod number per hill and soybean yield. Their correlations are likely to be negative with SD delay (Figure 6a,b). Such results support previous findings that abiotic/biotic stresses can adversely affect CGR with a similar mechanism for reducing soybean yield that is through decreases in nodes, pods [11], and seed number per plant [46], which are affected by CGR during vegetative and/or early reproductive periods [11,47]. Taken together, our results lead to a conclusion that the yield of double-crop soybean, which is liable to be late planting, might be determined mainly by how much growth has already been achieved before flowering and through R2–R5. In this study, soybean yield showed a significant reduction of 27.0% with an SD delay of 30 days (i.e., from S1 to S3). However, it did not show a significant reduction with an SD delay of 15 days (i.e., from S1 to S2). With respect to optimal SD without yield loss for double-crop soybeans, our data suggest that there is only a short time window (i.e., <30 days) in the southern area of Korea.

5. Conclusions

In this study site (35°10′ N), we were able to demonstrate that the flowering phenology of double-crop soybean cultivated in drained paddy fields could be accelerated by 5 days with each 15-day delay in SD within the range from 10–15 June to 10–15 July. More importantly, such a change in flowering phenology can lead to a reduction in vegetative growth (i.e., LAI, AGBM) up to R2 and subsequent reductions in overall growth (i.e., LAI, NAR, and CGR) and yield parameters (i.e., node, pod, and seed numbers) throughout crop stages after flowering, resulting in a 27.0% decrease in yield when SD was delayed by 30 days. The better performance of RT over DPRT was demonstrated by the greater yield of 13.7%, whereas they had an overall minor effect on crop phenology and growth. This further points to the fact that DPRT, which holds high volumetric soil water, teamed with extreme wet events could increase the risk of soybean yield reduction in drained paddy fields. Our data highlight that the yield of double-crop soybean cultivated in drained paddy fields was greatly varied with weather volatility between cropping years as well as SD and TS, ranging from 123.8 to 552.0 g m⁻² over three years. Our findings also demonstrate that, among variables considered, weather volatility between cropping years was the greatest contributor to yield variability (30.4%), followed by SD (17.0%) and TS (10.3%). Taken together, our results lead to the conclusion that the yield of double-crop soybeans is mainly determined by how much growth has already been achieved before flowering and through R2–R5. Our results also suggest that TS and SD might be important options to avoid or minimize the effects of adverse weather conditions. While we believe that the study presented here provides valuable data as to the production of double-crop soybeans grown in drained paddy fields, this study also has some limitations in terms of data applicability across an extensive area. Hence, further studies should also be added with field observations across a much more extensive area as well as with modeling approaches to enhance our ability to understand and/or minimize the uncertainty in the production of double-crop soybean cultivated in a wide range of drained paddy fields.