Environmental E ﬀ ects on Soybean ( Glycine Max (L.) Merr) Production in Central and South Germany

: The cultivation area of soybean ( Glycine max (L.) Merr) is increasing in Germany as a way to ensure self-su ﬃ ciency through its use as feed and food. However, climatic conditions needed for soybean cultivation are not appropriate in all parts of the country. The objective of this study was to determine the inﬂuence of solar radiation, temperature, and precipitation on soybean seed productivity and quality in central and south Germany. A multi-factorial ﬁeld trial was carried out with three replicates at four locations in 2016 and ﬁve locations in 2017, testing 13 soybean varieties from the maturity groups MG 00 and MG 000. Considering all the tested factors, “variety” was highly signiﬁcant concerning protein content (Ø 41.1% dry matter (DM)) and oil content (Ø 19.1% in DM), but not seed yield (Ø 40.5 dt ha − 1 ).The broad sense heritability of protein content was H 2 = 0.80 and of oil content H 2 = 0.7. Protein and oil content were signiﬁcantly negatively correlated (r = − 0.82). Seed yield was signiﬁcantly positively correlated with solar radiation (r = 0.32) and precipitation (r = 0.33), but signiﬁcantly negatively with Crop Heat Units (CHU) (r = − 0.42). Over both experimental years, varieties from maturity group MG 00 were less signiﬁcantly correlated with the tested environmental factors than varieties from maturity group MG 000. None of the environmental factors tested signiﬁcantly increased the protein or oil content of soybean. In growing areas with heat periods during ripening, protein content tended to be higher than in cooler areas; in areas with high solar radiation during ﬂowering, protein content tended to be reduced.


Introduction
Soybean (Glycine max (L.) Merr.), with its multiple uses, is one of the most important crops worldwide [1]. The high-quality of its protein makes it an important component of animal feed for both milk and meat production. Soybean oil is also used for human nutrition. As the demand for vegetarian and vegan food has increased, vegetable proteins and oils from soybeans have gained importance in recent years [2]. The world soybean production in 2017 was 352.6 million tons, of which 77% was GMO soybean [3,4]. The average soybean grain yield in the entire European Union in the years 2016-2017

Cultivation
The field experiments were conducted in two years and at four locations in Germany in 2016-Gruenseiboldsdorf: GSD; Eckartsweier: EKW; Wetterau: WET; Guesten: GUS; and five locations in 2017-Added Rossleben: ROS (Table 1). The soil texture at all locations was sandy loam with optimal pH 6.5-6.9. The experimental design at all locations was a randomized complete block design with three replicates and 13 varieties. The plot size was identical with harvest area ( Table 1).
The seeds of all varieties were untreated and inoculated with HiStick ® Soy (2 × 10 9 colony forming units per gram nitrogen fixing bacterium Bradyrhizobium (Bradyrhizobium japonicum); the application rate was 400 g per 100 kg of seeds). The seeding depth was 3-5 cm and seed rate was calculated to achieve plant density of 60-65 plants m −2 . In all years and at all locations, weeds and pests were controlled chemically according to the best management practices. For weed control at the GSD, WET, and EKW locations, the mixture of Centium ® 36 CS (360 g l −1 Clomazone) 0.25 l ha −1 + Sencor WG (700 g kg −1 Metribuzin) 0.3 kg ha −1 + Spectrum (720 g l −1 Dimethenamid-P) 0.8 l ha −1 was applied as pre-emergence weed control. In the GUS location Artist ® (240 g kg −1 Flufenacet, 175 g kg −1 Metribuzin) 2.0 l ha −1 + Centium ® 36 CS (360 g l −1 Clomazone) 0.2 l ha −1 were additionally used in pre-emergence. In ROS, chemical weed control consisted of pre-emergence spraying of Stomp ® Aqua (455 g l −1 Pendimethalin) 2.4 l ha −1 + Spectrum (720 g l −1 Dimethenamid-P) 1.4 l ha −1 . In 2016 at all locations Karate Zeon ® (9, 4%) Lambda-Cyhalothrin was applied at 0.075 l ha −1 against butterfly painted lady (Vanessa cardu). In the second trial year, there was no infestation by this insect. No fungicides were used, as none are officially registered for soybeans in Germany and none of the diseases exceeded the threshold.

Data Collection
Seed yield was determined by combine harvesting at full maturity (R8 (full maturity) + 8-15 days) [31]. The water content of seeds was measured immediately after harvest, and seed yield per plot was corrected to 14% moisture in seed. Protein yield and oil yield were calculated by multiplying seed yield with their corresponding protein or oil content. From all plots at all locations, samples were taken (400-500 g intact soybean seeds) to determine the protein and oil content via near-infrared reflectance spectroscopy (NIRS, Polytec PSSSHA03-2.1) [32].
Weather data was provided by the German weather service. The weather stations were a maximum of 5 km away from the trial fields (Table 3). The total precipitation over the entire vegetation period (Pwv) was summed for each month, for the growing season (15 April-15 September), and for the critical period for drought stress (15 June to 15 August, R1-R6-Flowering-seed filling). The sum of solar radiation (SRwv) was calculated from the daily value in kWh m −2 for the period from 15 April to 15 September in all trials and years. Solar radiation during flowering (SRf) is the sum of daily value solar radiation in kWh m −2 from 1 June to 1 July. The sum of CHU for the entire vegetation period (CHUwv) was calculated according to Rossberg and Recknagel [15] for the period from 1 June to 15 September. Crop Heat Units (CHU) at maturity was determined for the period 15 August to 15 September. Temperature (T night ) at flowering is the average of temperatures for every night from R1 to R3 (15 June-31 July).

Data Analysis
We first calculated the mean of each individual variety for each trait in the respective environments. The means of individual varieties were then averaged across the environments by calculating the arithmetic mean. The normality of distribution and homogeneity of variance of residuals were checked graphically.
For computing broad sense heritability, we estimated the genetic variance in a full random model Equation (1) based on the variety means of each single environment. The standard error of entry means was estimated in a model that considers interactions with the variety as random and all other effects as fixed Equation (2). Analyses were carried out with the software R [33] using the packages lmerTest [34], lsmeans [35], and lme4 [36].
Equation (1): where P ikl is the observed phenotype of the ith variety in the kth year at the lth location, G i is the random effect of the ith variety, Y k the random effect of the kth year and L l the random effect of the lth location, (GL) il is the random effect of the ith variety in the lth location, (GY) ik is the random effect of the ith variety in the in the kth year, and (YL) lk is the random effect term of the lth location in the kth year. e ikl denotes the error term. Equation (2): where P ikl is the observed phenotype of the ith variety in the kth year at the lth location, g i is the fixed effect of the ith variety, y k the fixed effect of the kth year and l l the fixed effect of the lth location.
(GL) il is the random effect of the ith variety in the lth location, and (GY) ik is the random effect of the ith variety in the in the kth year. Fixed effects are written in bold lowercase letters. e ikl denotes the random error term. Based on the genetic variance (1) and standard error (2) broad sense heritability was computed as described in (3) [37].
Equation (3): where σ 2 G is the genetic variance derived from the full random model (1) and SE 2 is the squared standard error of the difference between the means, derived from (2).
Interrelationships between environmental impacts and seed properties were determined by calculating the Pearson coefficients of correlation (r).

Results
The analysis of variance showed highly significant effects of variety on protein and oil content ( Table 4). The interaction of year × location was also highly significant for protein and oil content.  (1)) and heritability (Equation (3)) for seed yield and quality over five locations and two trial years. The highest heritability was estimated for protein content with H 2 = 0.80. The heritabilities for seed and protein yields were H 2 = 0.55 and for oil yield H 2 = 0.61.
The mean seed yield of varieties from maturity group MG 00 was higher than that from maturity group MG 000 ( Table 5). The means of protein and oil content were almost equal for both maturity groups. The responses of varieties grouped by maturity groups and depending on year revealed significant correlations ranging from r = 0.25 to r = 0.74 (Table 6). In Table 6, the effects of year are shown. Varieties from the maturity group MG 000 were significantly correlated with environmental conditions compared to varieties from the maturity group MG 00. Solar radiation was the main factor leading to an increase in seed yield and thus to increases in protein yield and oil yield. Precipitation often significantly increased yield and thus protein yield and oil yield in MG 000 varieties. Crop Heat Units, which had no significant effects on the observed traits in MG 00 varieties, were negatively correlated with seed yield, protein yield, and oil yield in MG 000 varieties. The response of MG 000 varieties to precipitation correlated positively with seed yield, protein yield, and oil yield. Crop Heat Units (CHU) and CHU at maturity correlated negatively with yield and thus protein and oil yields.
There was a strong negative correlation (r = −0.82 ***) between oil content and protein content over all varieties, locations, and trial years (Figure 1a). Varieties from the maturity group MG 000 were significantly correlated with environmental conditions compared to varieties from the maturity group MG 00. Solar radiation was the main factor leading to an increase in seed yield and thus to increases in protein yield and oil yield. Precipitation often significantly increased yield and thus protein yield and oil yield in MG 000 varieties. Crop Heat Units, which had no significant effects on the observed traits in MG 00 varieties, were negatively correlated with seed yield, protein yield, and oil yield in MG 000 varieties. The response of MG 000 varieties to precipitation correlated positively with seed yield, protein yield, and oil yield. Crop Heat Units (CHU) and CHU at maturity correlated negatively with yield and thus protein and oil yields.
There was a strong negative correlation (r = −0.82 ***) between oil content and protein content over all varieties, locations, and trial years (Figure 1a). Protein content ranged from 35.4% to 45.1% with a mean of 41.1%. Oil content ranged from 17.1% to 21.6% with a mean of 19.1% (Table 5). The correlation between protein yield and oil yield was positive at r = 0.84, at the highly significant level (p < 0.01) (Figure 1b).
The correlation between seed yield and protein content was low and not significant (data not shown). In contrast, seed yield and oil content were significantly negatively correlated at r = −0.28 *** (Figure 2). Protein content ranged from 35.4% to 45.1% with a mean of 41.1%. Oil content ranged from 17.1% to 21.6% with a mean of 19.1% (Table 5). The correlation between protein yield and oil yield was positive at r = 0.84, at the highly significant level (p < 0.01) (Figure 1b).
The correlation between seed yield and protein content was low and not significant (data not shown). In contrast, seed yield and oil content were significantly negatively correlated at r = −0.28 *** (Figure 2).

Discussion
The high heritability of traits contributes to genetic progress, a characteristic that informs the selection of the most suitable genotypes as future varieties [17,38]. The conditions under which soybeans are grown, including the environment, determine the breeding direction, and this in turn depends on the geographic regions and the proposed cultivation of specific varieties [39,40]. Weather conditions differed in both trial years (Table 3); only night temperatures and solar radiation did not vary greatly over all locations. On average, the year 2017 was moister and cooler than 2016. Crop Heat Units at maturity in 2017 were approximately 60-160 units lower than in 2016. Solar radiation at flowering and-seed filling however, was on average 30 kWhm −2 higher in 2017 than in 2016. As a result, the soybean production properties correlated with environmental factors with varying degrees of intensity ( Table 6).
The most critical period in the entire vegetation period of soybean concerning water availability and sufficiency is R4-R6 (pod formation and seed filling) [14,18,21,41]. The results of the current study, with a positive correlation (r = 0.47) between seed yield and precipitation during floweringseed filling are in accordance with existing literature. The early maturing varieties of MG 000 were more sensitive to increases in precipitation in the period R4-R6 than MG 00 varieties in each trial year ( Table 6). This may have been due to the shorter seed filling period and shorter duration of individual periods of vegetation growth of soybean from maturity group 000. This agrees with Souza et al. [42], who compared MG 00 (early ripening) and II (later ripening) varieties in Brazil and also with Chen und Wiatrak [41], who performed experiments with MG IV to VIII in the USA. High solar radiation at flowering increased seed yield in the current study over the two experimental years independent of the maturity group (Table 6), similar to findings of other researchers [13,39,43]. The intensification of solar radiation at the time of flowering could stimulate photosynthesis and thus raise the production of assimilates, leading to an increase in seed yield by increasing the number of pods per reproductive node [39]. Similar results (r = 0.37) for five maturity groups in the USA have also been reported by Kane und Grabau 1992 [44]. At the same time, high temperatures at certain stages of development (R3-R6) can have a negative impact on seed yield and quality characteristics of soybean. For example, Dornbos and Mullen and Gibson and Mullen [45,46] reported that high temperatures (>29 °C) led to a reduction in number of seeds per plant and in seed weight The negative correlation between CHU at maturity and seed yield in our study (r = −0.42) was higher for the early maturing

Discussion
The high heritability of traits contributes to genetic progress, a characteristic that informs the selection of the most suitable genotypes as future varieties [17,38]. The conditions under which soybeans are grown, including the environment, determine the breeding direction, and this in turn depends on the geographic regions and the proposed cultivation of specific varieties [39,40]. Weather conditions differed in both trial years (Table 3); only night temperatures and solar radiation did not vary greatly over all locations. On average, the year 2017 was moister and cooler than 2016. Crop Heat Units at maturity in 2017 were approximately 60-160 units lower than in 2016. Solar radiation at flowering and-seed filling however, was on average 30 kWhm −2 higher in 2017 than in 2016. As a result, the soybean production properties correlated with environmental factors with varying degrees of intensity ( Table 6).
The most critical period in the entire vegetation period of soybean concerning water availability and sufficiency is R4-R6 (pod formation and seed filling) [14,18,21,41]. The results of the current study, with a positive correlation (r = 0.47) between seed yield and precipitation during flowering-seed filling are in accordance with existing literature. The early maturing varieties of MG 000 were more sensitive to increases in precipitation in the period R4-R6 than MG 00 varieties in each trial year ( Table 6). This may have been due to the shorter seed filling period and shorter duration of individual periods of vegetation growth of soybean from maturity group 000. This agrees with Souza et al. [42], who compared MG 00 (early ripening) and II (later ripening) varieties in Brazil and also with Chen und Wiatrak [41], who performed experiments with MG IV to VIII in the USA. High solar radiation at flowering increased seed yield in the current study over the two experimental years independent of the maturity group (Table 6), similar to findings of other researchers [13,39,43]. The intensification of solar radiation at the time of flowering could stimulate photosynthesis and thus raise the production of assimilates, leading to an increase in seed yield by increasing the number of pods per reproductive node [39]. Similar results (r = 0.37) for five maturity groups in the USA have also been reported by Kane und Grabau 1992 [44]. At the same time, high temperatures at certain stages of development (R3-R6) can have a negative impact on seed yield and quality characteristics of soybean. For example, Dornbos and Mullen and Gibson and Mullen [45,46] reported that high temperatures (>29 • C) led to a reduction in number of seeds per plant and in seed weight The negative correlation between CHU at maturity and seed yield in our study (r = −0.42) was higher for the early maturing varieties (MG 000). MG 000 varieties can mature in the middle of September in Germany, whereas MG 00 varieties are usually harvested 10-12 days later. High temperatures during the final period of ripening may interfere with the physiological maturing process and simply lead to desiccation, not perfect maturing [9,47].
As previous studies suggest, the protein content of soybean seeds is influenced by day and night temperatures and varies depending on their stage of development [18,47]. Temperatures of about 20 • C correlated significantly positively with soybean protein content. Our results indicate that higher temperatures at maturity may lead to higher protein content ( Table 6). The protein content of soybean seeds increased with increasing mean temperatures >28 • C and remained constant <28 • C [18]. Compared to our studies, previous studies quoted above examined varieties with a much longer vegetation period (MG up to VIII) and in locations at a much lower latitude (<29.4 • S). In our study, the early ripening varieties (MG 00 and MG 000) reacted more rapidly to the higher temperatures they experienced, which have begun to occur more often in Germany.
Oil content was the only seed property that correlated negatively with T night at development stage R1-R3, similar to previous results [48][49][50]. However, soybean studies at 23 locations in Argentina indicated the opposite [27]. Such differences in results were likely due to the fact that those studies [25] were conducted in different regions in Argentina while other three were performed in the USA (Iowa) [48][49][50]; the varieties examined, therefore, belonged to different maturity groups (MG IV-V) (middle late) and II (late).
The time during the growing season when drought occurs can affect the intensity of its influence on protein and oil content [21]. It is well established from studies in the USA, South America, Central Europe, and Russia that oil content is negatively correlated with precipitation at flowering-seed filling phase [14,20,23,47]. Our studies, conducted in a region (Germany) previously considered not suitable for soybean cultivation (because only few varieties were able to ripen at the low CHU), is in line with these results (Table 6). A much stronger negative correlation, as compared to the average of both years, was observed in the dry year 2016 compared to 2017 (Table 6). It is possible that if the season were already dry, further water deficits at this stage could have had a stronger increasing effect on the oil content. If water deficiency occurred at R5-R6, soybean seeds did not fully develop and did not reach their variety-specific size [42,49,51,52]. The transfer processes of nutrients (first of all carbohydrate) and thus seed quality were impaired [14,47], and the assimilates could not be completely converted into protein and oil [52]. However, the protein and oil content are negatively correlated, so the protein content of soybean seeds increased by 3-5% under drought stress then oil content decreased by 2-3% compared to seeds of control plants with optimal water supply [48]. Protein formation in soybean seeds starts about 10-15 days earlier than that of fats [20]; when drought becomes more severe, protein synthesis continues but nutrients for oil are exhausted [21]. From a biochemical point of view, this can be linked to the activation of superoxide, which, in a drought, promotes the accumulation of protein more than that of oil [53]. Such a phenomenon is already known from cereals; when under drought stress, starch enzyme activity is more strongly affected than enzyme for protein formation [54]. Therefore, at drier sites and/ or at locations where drought periods can be expected, varieties that have been selected for high oil content and oil yield should not be grown; rather, this production objective should be pursued at more humid locations. The negative correlation between precipitation at flowering-seed filling and oil content was only detected under dry conditions; under more humid conditions we could not detect this effect (trial year 2017).
Negative correlations of oil content with protein content of soybeans have been reported for soybean since the mid-1960s [17]. The highest negative correlation between protein and oil content was found in the European early maturing genotypes (0-0000); for example, almost r = −0.9 in Germany [17], r = −0.79 in Austria [23], and r = −0.83 in Slovenia [55]. According to investigations in Russia over 27 years, this correlation was on average r = −0.53 [20] and in the Midwest of the USA r = −0.64 [56]. Piper und Boote analyzed 1863 varieties from 10 maturity groups at 60 locations throughout the USA and documented a correlation of r = −0.43 [18]. This strong negative correlation between protein content and oil content was also seen in our results r = −0.82 (average over only early maturing MG 00 and MG 000 varieties- Figure 1a), as typical for Central Europe. The reasons for the strong negative correlation between protein and oil content are not completely clear. These two properties are determined by the same combination of genes and it is difficult to do plant breeding work only to increase of one of these properties without leaving the other unchanged [20,57,58]. However, the synthesis of protein and oil in soybean seeds utilize the same carbon metabolism, and the synthesis processes interact with each other [59]. Therefore, competition for carbon can develop, and this is reflected in a negative correlation.
If the aim is to increase total soybean protein production in Germany, increasing seed yield would increase the amount of protein per unit area. In this context, the phenotypic correlations of seed yield with protein and oil content are important. Though seed yield correlated negatively with protein content but positively with oil content in the common studies [17,56], our investigation did not find a significant correlation of seed yield with protein content; this is similar to the findings of Whaley und Eskandari [40]. Despite its importance, there is no clear evidence of a relationship between seed yield and protein content. Test results from Germany provide information about a low and negative correlation [17] while in the USA a low but positive correlation has been calculated [22]. This correlation is influenced by year and weather/soil conditions in the growing area [23]. Such a regulation process between protein content and seed yield is known from wheat [60]. Therefore, genotypic variation together with environment provides many breeding possibilities for the adaptation of new varieties [40]. The low negative correlation between seed yield and oil content in our study contrasts with other available results [22,51]. It is essential to take the findings of referenced studies into consideration when planning the production, especially the choice of varieties, of certain soybean products such as soybean meal, soybean oil or tofu, and other foodstuffs, in different growing areas of a country with a temperate climate.
Time to maturity, drought tolerance, and heat tolerance are important traits to consider in adapting soybean varieties to different growing environments. Since the levels of protein content and oil content of the maturity groups MG 00 and MG 000 barely differed, and since the seed yields of MG 00 varieties trended higher than those of MG 000 varieties, MG 00 varieties should preferentially be cultivated and bred in regions suitable either for protein or oil production. To select for varieties with high seed yield and with long ripening period, such as the tested MG 00 varieties, the testing should occur in regions with comparatively high solar radiation (for example location Eckartsweier). The late ripening varieties from MG 0 would also be appropriate for these locations. In regions with comparatively high precipitation, MG 000 varieties would be suitable because of their high seed yields and therefore high protein and oil yields (for example Grünseiboldsdorf, Wetterau, and Rossleben). If MG 00 varieties were cultivated in drier areas (such as Eckartsweier and Guesten) the oil content would be higher. High temperatures at maturation would result in higher protein content in MG 000 varieties (Eckartsweier, Wetterau and Grünseiboldsdorf). The following table ranks the tested soybean varieties according to their adaption to locations with different environmental properties in Germany (Table 7). Table 7. Classification of soybean varieties according to their adaption to different locations in Germany.

Location MG 00 Varieties MG 000 Varieties
Suitable for High dry oil content oil content wet no effect yield warm yield protein content cool no effect no effect high solar radiation yield yield

Conclusions
Environmental parameters such as solar radiation, temperature, and precipitation, in addition to breeding progress, will determine the success of soybean production in Germany. It is necessary to place the varieties in suitable growing environments according to the direction of use to achieve the highest performance. Regions with high solar radiation and ample precipitation are suitable for producing high seed, protein, and oil yields of soybean using varieties from MG 00 and MG 000 maturity groups. Dry growing areas should be selected for oil production as a possible supply of vegetable oil, especially in organic farming. If production is concentrated on high protein content, the MG 000 varieties should be grown in regions where temperatures remain high during ripening. This knowledge can be used as a tool to expand soybean cultivation to the dry eastern Germany and northern German cool regions. Domestically grown soy products can then meet the growing demand for vegan and vegetarian food.