Plant material

Seeds of three maize (Zea mays L.) inbred lines derived at the Maize Research Institute, Zemun Polje (G1-217, G2- G325/75-2 and G3-G335/99) were used in the trial. One inbred (G1) belonged to the FAO maturity group 400, while the remaining two (G2, G3) belonged to the FAO maturity group 600.

Soil properties

The soil was a slightly calcareous Chernozem, namely Molcal silt loam (coarse-loamy, mixed, superactive, mesic Vitrandic Calcixerolls) (^{USDA-NRCS, 1999}). The 0 to 30 cm layer contained 3.3% organic matter, 0.21% total N, 1.9% organic C, 14 mg available P 100 g^-1 soil, 31 mg extractable K 100 g^-1 soil, 9.7% total CaCO₃, and pH 7.8.

Seed production and processing

The 2-yr trial was carried out according to the split-plot system in the location of Zemun Polje (44°52'00" N, 20°19’00” E) in the vicinity of Belgrade, Serbia, in the period from 2018 (Y1) to 2019 (Y2). Sowing was performed on the following five dates: 1 April (T1), 10 April (T2), 20 April (T3), 30 April (T4) and 10 May (T5).

Seeds for sowing were divided by size into three fractions by passing through sieves of the appropriate mesh size with the aid of the Carter Day laboratory maize processing equipment (CEA, Minneapolis, Minnesota, USA). A small seed size (SD) amounted to 6.5-8.4 mm, while large (LD) and primary (PD) seed sizes amounted to 8.5-11.0 and 6.5-11.0 mm, respectively. The elementary plot size amounted to 4.0 m × 6.3 m. The plot consisted of nine rows with one inbred (3×3 rows of each seed size - SD, LD, PD). The inter-row distance was 0.7 m and the plant distance was 0.2 m (hence, the plant density was 71 000 plants ha^-1). Each trial variant was sown in four replicates for each inbred. The trial was performed under rainfed conditions (Table 1).

Table 1 Mean monthly temperatures. 

Y1: 2018; Y2: 2019; Y3: reference period (1981-2010).

Laboratory testing

Physical properties of maize seeds, width (W), length, (L), thickness (Tk), and the ratio of seed fractions on the cob (small fraction, SF/large fraction, LF), were evaluated under laboratory conditions after harvest of inbreeds.

After harvesting and drying of ears, samples were drawn to determine percentage ratio of large (LF) to small seeds (SF). Five ears each were selected and shelled and then seeds were passed through sieves of mesh sizes 6.5-8.4 mm (SF) and 8.5-11.0 mm (LF). Sifted seeds were weighed to determine the proportional ratio of seeds by size.

The seed W, L and Tk were measured with a calliper (accuracy 1/10 mm, range 0-150 mm; Kern, Balingen, Germany) from five-seed samples drawn from the central part of the ear in three replicates.

The determination of the grain yield was done by measuring seed weight from middle rows of all maize ears, and the estimation was done at 14% moisture.

Meteorological data

The climate of Serbia can be described as temperate-continental with more or less local characteristics, which correspond to the climate of South-eastern Europe, the Balkans.

According to the Republic Hydrometeorological Service of Serbia (^{RHMS, 2021}), temperatures in 2018 and 2019 were the highest since the beginning of recording temperatures in Serbia (1888). The average annual temperatures in relation to the reference period (1981-2010) deviated by 2.0 °C for 2018 and 2.2 °C for 2019. Mean monthly temperatures in the April-August period were characteristic of hot springs and summers (Table 1).

The precipitation distribution under the conditions of continental climate is uneven, with higher amounts in the warmer half of the year (April-September) and the annual precipitation sum ranges 540 to 820 mm. Maize is a plant with increased needs for water, particularly in the stage of intensive growth. The critical period in which the plant needs a sufficient moisture is 15-20 d after tasselling and 10-15 d after pollination. The precipitation sum during the 2-yr trial was almost equal for Y1 (336.5 mm) and Y2 (366 mm), which was lower by 130 and 100 mm, respectively, than the sum in the reference period (1981-2010). The highest differences occurred in April and May in Y1 at the time of sowing (Table 2).

Statistical analysis

The descriptive statistics was applied to process each of obtained parameters. Differences among different sowing dates were determined by the ANOVA and the LSD test at the 5% and 1% risk levels. In order to draw conclusions objectively about the impact of observed factors on studied traits of maize seeds, parametric tests (ANOVA and LSD-test), Person’s correlation coefficient, linear regression analysis (Y = β₀ + β_1xi + ε) and the principal component analysis (PCA) were applied. The obtained experimental data were processed using the statistical package SPSS 21 (free version, IBM, Armonk, New York, USA).

Yield in relation to seed physical properties

By the observation of factor loads for the seed size traits (W, L, Tk, SF, LF) and GY, PCA analysis shows that two components represented 88.28% of the obtained variation. Length, Tk, SF, LF and GY were positively correlated with the first component (73.37%). On the other hand, the physical seed trait W was correlated with the second component (15.91%) (Figure 3b). By observing the distribution of regression factor score values of the research variant sowing date + seed sizes at sowing, it was determined that the most of the coefficients were located on the right side of the diagram as well as the component matrix of physical seed properties (Figure 3a). It can be concluded that the first component was a set of cropping practices that determined the formation of physical seed properties and yield. Late sowing, T5, was negatively correlated with all properties (W, L Tk, SF, LF) and GY. The second component, which was far less efficient than the first, was seed quality at sowing, responsible for W variability (0.964).

All properties of the seed size participate with over 50%, r ≥ 0.500, in the yield variance (Table 6). Across environmental and genotypic sources of variation, the grain yield is more closely related to the seed number than to the seed weight (^{Dokic et al., 2020}). The variation in seed weight is genetically determined, but it is often a result of adjusting to the number and shape of seeds.

The seed size variability mainly affected the grain yield of the genotype G1 (R² ≥ 0.600), while variability in seed W, L and Tk had lesser effect on genotypes G2 and G3 (0.24 ≤ R² ≥ 0.50).

The highest effect of variability in seed morphological traits on GY was related to the sowing date. The seed size had nonsignificant effect on yield variability related to the first four sowing dates (R² ≤ 0.10). On the other hand, all seed size properties (W, L, Tk) on T5, equally affected GY (R² ≥ 0.580).

The effect of the year of production on the seed size to GY ratio was expressed in Y2, in which the yield was determined by the variability in the seed size (R² ≥ 0.49). It varied the most with the change of seed L (R² ≥ 0.60). In Y1, the seed size did not affect the yield.

SF: Small fraction; LF: large fraction; SD: small seeds 6.5-8.4 mm; LD: large seeds 8.5-11.0 mm; PD: primary seed sizes 6.5-11.0 mm; SDT1: seeds of small size sown on 1 April; LDT1: seeds of large size sown on 1 April; PDT1: seeds of primary size sown on 1 April; SDT2: seeds of small size sown on 10 April; LDT2: seeds of large size sown on 10 April; PDT2: seeds of primary size sown on 10 April; SDT3: seeds of small size sown on 20 April; LDT3: seeds of large size sown on 20 April; PDT3: seeds of primary size sown on 20 April; SDT4: seeds of small size sown on 30 April; LDT4: seeds of large size sown on 30 April; PDT4: seeds of primary size sown on 30 April; SDT5: seeds of small size sown on 10 May; LDT5: seeds of large size sown on 10 May; PDT5: seeds of primary size sown on 10 May.

Figure 3 Principal component analysis of the variables in the study concerning effects of sowing date (T1-T5) and seed size on physical properties width (W), length (L), thickness (Tk) and grain yield (GY) (a) and eigenvectors of principal component, sowing date and seed sizes at sowing (SD, LD, PD) (b). 

Table 6 Pearson’s coefficient of correlation. 

*Significant correlation among seed physical properties (width, length, thickness) according to the F-test (P < 0.05).

The yield varied mostly when SF seeds were sown. Within this fraction, all seed size properties equally contributed to the variation of GY, R² ≥ 0.40.

There were nonsignificant differences in the obtained yields related to the first four sowing dates. The lowest yield and the highest variability σ = 8.029 were obtained with the latest sowing date, T5. The sowing date determined the weather conditions to which maize is exposed. The sowing date was determined by the weather conditions to which maize was exposed. Variations in weather conditions associated with the sowing date were the main factors influencing the maize growth. The significant reduction in yield occurred due to late sowings because of the reduced post-silking DM production (^{Zhou et al., 2016}). An interesting approach in interpreting the key role of the seed size in determining the number of seeds per plant is that the seed size plays a significant role in modulating genetic and external control of the seed number.

The sowing dates that contribute to a better relationship between the seed size and the yield is the consistency of drying temperature, duration of the treatment and biological effect of the process (^{Egli and Rucker, 2012}).

Late sowing reduces the effective grain filling rate compared to early sowing. The plant growth rate in the grain filling stage was slowed down in late sown plants due to a low level of daily radiation and radiation use efficiency. Daily temperatures significantly affected the grain filling rate increasing the weight by 0.3 mg per day per °C. Furthermore, there was a reduction in the source of assimilative, which caused the reduction of the seed weight (^{Kanatas et al., 2020}). Studies aimed at more complete understanding, estimation and evaluation of the properties of hybrid maize seeds provide the correct perception of new production conditions due to changes that occur in the form of temperature extremes and water deficits. Under the impacts of continental climate, weather can act as a strong abiotic factor affecting the maize seed yield (^{Li et al., 2019}; ^{Ghete et al., 2020}). Climatic parameters such as water availability and temperature levels have been identified as key factors responsible for maize crop productivity (^{Tokatlidis, 2013}; ^{Butts-Wilmsmeyer et al., 2019}). The precipitation sum in the April-August period affects maize yields more than N fertiliser does (^{Halmajan et al., 2017}). Maize is particularly susceptible to drought stress during pollination and grain filling (^{Wang et al., 2019}; ^{Iqbal et al., 2020}).

In the conducted studies, there were several groups of factors: controlled, partially controlled and absolutely uncontrolled ones. The genotypic combination and crop sowing were completely controlled factors. The seed size was a partially controlled factor, while agro-ecological conditions and the production location were absolutely uncontrolled factors. The degree of effects of the stated factors varied. Effects of sowing dates were just of partial importance. Such pattern of expression of sowing dates was contributed by ecological conditions in the period of studying. High temperatures (18.0-22.7 °C) at sowing (April-May, in Y1) and low levels of soil moisture (24.6-39.0 mm) resulted in a delay of germination, and thus the effect of sowing intervals was reduced. The fifth sowing date, T5, caused the greatest variations. The effect of ecological conditions at sowing on the seed size was even over 70% (R² > 0.70). The yield variance for T5 amounted to σ = 8.029. Later sowing dates coincided with high temperatures and drought. Moreover, T5 unfavourably affected stages of pollination and grain filling. In Y1 poor pollination was a result of high percentage of large fraction seeds (LF > 70%), and in Y2 poorer grain filling and lower GY.

A comparison of weather parameters across years of investigation showed that the daily mean temperature was the highest during most of the flowering stage. This stage is the most susceptible stage of the maize growth in terms of the impacts of high temperatures (^{Sharon et al., 2013}; ^{Hatfield and Prueger, 2015}; ^{Vinayan et al., 2020}), causing significant losses in grain yields, mainly by affecting pollen viability (^{Windhausen et al., 2012}; ^{Alam et al., 2017}). The variability of the seed size indicated that production conditions had the greatest impact within the hybrid combination and then within the growing practices, particularly sowing dates (^{Nave et al., 2016}). The influence of the seed size at sowing was small and not so significant for the formation of good morphological properties of seeds, but it was significant for the creation of conditions for uniform germination that was a prerequisite for stable yields.

It is more than likely that the temperature and precipitation will increase in the near future, as projected in most weather modelling studies. However, the increase in precipitation will not compensate for the water loss on the surface of the planet due to intense thermal variations, which will increase the significance of irrigation. In addition, changes in heat resources due to regional warming will accelerate the crop growth. It has been predicted that the period of maize growth, especially in the grain filling stage, would be shortened by about 10 to 30 d in the current maize growing regions, which would lead to reduced yields (^{Wang et al., 2011}).

The results of the simulation of climate changes in the temperate continental climate indicate that agriculture cannot avoid effects of climate changes, and unfortunately most of these effects will be adverse (^{Fodor and Pásztor, 2010}).