Modelling the growth, development and yield of Triticum durum Desf under the changes of climatic conditions in north-eastern Europe

Climatic conditions and phenology

The growth and development of T. durum plants was moderately differentiated by weather conditions in the analyzed years (Table 1).

The growing seasons of 2015, 2016 and 2017 lasted 136, 132 and 145 days, respectively; the sum of temperatures was determined at 2011.3, 1895.6 and 2069.9 °C, respectively, and the sum of precipitation was determined at 366.7, 360.1 and 350.5 mm, respectively. However, a comparison of cumulative temperatures and precipitation in the phenological phases of T. durum in each year of the study indicates that temperature and precipitation could have influenced the duration of the examined phases and plant growth indicators (Fig. 1). Weather conditions were generally favorable for the growth and development of T. durum in 2015 and 2016. Cumulative temperatures and precipitation were quite similar in 2015 and 2016 up to the booting stage, but precipitation levels in successive stages were higher in 2016 than in 2015. The growing season was shortest in 2016 and longest in 2017, mainly due to low temperatures during sowing and seed germination, and high precipitation during tillering, grain formation and ripening.

Biophysical parameters: LAI and SPAD

The LAI denotes the area of photosynthetic tissue per unit ground surface area (m² m⁻²). The LAI is directly associated with plant canopy, and it is an indicator of net primary production, water and nutrient use, and the carbon balance. SPAD is a measure of leaf greenness that is directly associated with chlorophyll content and nitrogen sufficiency.

The main effects of LAI and SPAD were analyzed separately in the framework of the Zadoks scale to reveal the significant effects of years, nitrogen rates and sowing density, and an absence of significant effects associated with the application of the growth regulator (see Tables 1.1–1.6 in the Supplementary Information). In the analyzed years, LAI and SPAD were similar in the 2nd node detectable stage (Z32), but they differed in the stem elongation stage (Z45) and the ear emergence and heading stage (Z59), when LAI values were higher and leaf greenness values were lower in 2016 and 2017 than in 2015. These findings can be attributed to moderate temperatures and precipitation in 2015, and high precipitation in the critical growth stages in the remaining years. The general trend associated with the nitrogen rate was similar across the examined growth stages, i.e. a significant increase in LAI and SPAD values with nearly identical effects were noted in treatments with nitrogen rates of 80 and 120 kg ha⁻¹. A similar trend was observed in sowing density. In treatments with a sowing density of 450 and 550 germinating seeds per m², LAI values continued to increase, whereas SPAD values were below those noted in the treatment with a sowing density of 350 germinating seeds per m². The only significant interaction was observed between years and nitrogen rates.

The average values of LAI continued to increase in successive growth stages and were determined at 1.30 at Z32, 1.75 at Z45, and 1.99 at Z59. In turn, leaf greenness was significantly lower in the stem elongation stage (Z45) than in the preceding (Z32) and subsequent (Z59) stages.

The significant effect of the years × growth stages interaction for LAI and SPAD values resulted from similar means in stage Z32 in all years, as well as higher LAI values and lower SPAD values in subsequent growth stages in 2016 and 2017 than in 2015. In 2015, the increase in the nitrogen rate induced only a rising trend in LAI and SPAD values, whereas significant differences were observed in 2016 and 2017. To summarize, it should be noted that in successive Zadoks growth stages, the interactions between years, nitrogen rates and sowing density exerted significant effects on LAI and SPAD values, whereas the effects of year × nitrogen rate interactions were significant only in selected growth stages.

Contribution of different sources of variation to physiological and biophysical parameters of plant growth

The calculated eta-squares η² provide information about the contribution of different sources of variation to physiological variables (Table 2). The experimental years and agronomic factors (33.1% and 38.6%), growth stages, and interactions with other factors (32.5% and 39.3%) and random factors (34.4% and 22.1%) made similar contributions to the variation in the LAI and chlorophyll content. The variation in the net photosynthetic rate was related mostly to variations across years (32.6%) and the interactions between growth stages and other factors (24.3%). The variation in the transpiration rate was attributed mostly to the main effects of growth stages (45.8%) and the year × growth stage interaction (16.1%). Instantaneous WUE was strongly determined by variation in agronomic factors and growth stages (22.3% and 21.1%, respectively).

It is worth noting that the variation in agronomic factors made a considerable contribution to the total variation in LAI (22.3%) and SPAD (11.8%), but only a marginal contribution to the net photosynthetic rate (0.4%) and transpiration (2.0%).

Photosynthetic indicators— net photosynthetic rate, transpiration rate, and instantaneous water use efficiency

The effects of the net photosynthetic rate (Pn), transpiration rate (E) and instantaneous WUE were highly differentiated in successive growth stages, and relatively small differences were noted for agronomic factors (see Tables 2.1–2.9 in the Supplementary Information). At the same time, the analyzed photosynthetic indicators differed in successive stages of growth. The net photosynthetic rate was similar in the 2nd node detectable stage (Z32) and the stem elongation stage (Z45) at 29.7 μmol CO₂ m^–2 s^–1, and it was 15% higher at the end of the heading stage (Z59) than in the preceding stages. The transpiration rate continued to increase by 60% on average in successive stages of growth and development, from 1.59 H₂O m^–2 s^–1 in stage Z32, to 2.52 mmol H₂O m^–2 s^–1 in stage Z45, and 4.06 mmol H₂O m^–2 s^–1 in stage Z59.

An analysis of the results noted in different growth stages across years revealed significant year × growth stage and growth regulator × growth stage interactions (Fig. 2).

The year × growth stage interaction resulted from differences in the rates of photosynthesis and transpiration in the analyzed growth stages across years. In 2015, the net photosynthetic rate was similar in the first two growth stages, and it increased by around 30% at the end of the heading stage (Z59). In 2016, the photosynthetic rate continued to increase in successive growth stages. In 2017, the net photosynthetic rate was around 10% higher in the 2nd node detectable stage (Z32) than in the stem elongation stage (Z45) and at the end of the heading stage (Z59). The transpiration rate increased significantly in successive stages of plant growth and development, and the only exception was noted in 2015, when the analyzed parameter was similar in stages Z32 and Z45. The WUE index was highest in stage Z32, and a significant interaction was noted due to the correlation between the net photosynthetic rate and the transpiration rate in the remaining stages. Water use efficiency was similar in stages Z32 and Z45 in 2015, and in stages Z45 and Z59 in 2016, whereas significantly lower values in successive stages of plant growth were noted in 2017.

The growth regulator was the only agronomic factor that induced significant differences in the net photosynthetic rate across the examined growth stages. Photosynthesis indicators were similar regardless of the application of the growth regulator, and significant interactions resulted mainly from varied disproportions between the end of heading and the stem elongation stage in treatments with and without the application of the growth regulator.

It should be noted that the interactions between growth stages and nitrogen rates and sowing density were not significant, which implies that the effects of the interactions between increasing nitrogen rates and sowing density on photosynthetic indicators in successive growth stages were similar to the average values of photosynthetic indicators in the corresponding growth stages (Supplementary Information).

Agronomic traits

The means for yield components and yield are presented in Tables 3.1–3.8 of the Supplementary Information. Stem length was differentiated by the nitrogen rate and nitrogen rate × year interaction. Nitrogen rates of 80 and 120 kg N per ha increased stem length by 11% and 13%, respectively, relative to the unfertilized control. The significant year × nitrogen rate interaction resulted from the fact that the nitrogen-induced increase in stem length was smaller in 2015 (0.07 cm per 1 kg of nitrogen) than in 2016 and 2017 (0.09 cm per 1 kg of nitrogen). In 2015, ear length was similar to that noted in the remaining years, and only in 2017, ear length was 7% higher than in 2016. Ear length and the number of kernels per ear increased with a rise in nitrogen rate and decreased with a rise in sowing density.

Grain weight per ear and 1000 kernel weight were highest in 2015 and significantly lower in the following years. Grain weight per ear increased only in response to the nitrogen rate of 120 kg ha⁻¹, but 1000 kernel weight was not affected. Both traits decreased with a rise in sowing density. The significant year × nitrogen rate and year × sowing density interactions for both traits can be largely attributed to the magnitude of differences between years, rather than an increase or a decrease in this trend.

The biological yield (grain and straw) differed across years and nitrogen rates. In 2016, the biological yield was similar to that noted in 2015 and significantly higher (by 30%) than that noted in 2017. The biological yield increased by 28% and 35% in response to nitrogen rates of 80 and 120 kg ha⁻¹, respectively, relative to the unfertilized control. The significant year × nitrogen rate interaction was associated with variations in nitrogen use efficiency, and the difference between maximal biological yield was determined at 0.5 t ha⁻¹ in 2015, 2.3 t ha⁻¹ in 2016, and 2.8 t ha⁻¹ in 2017.

Grain yield was similar in 2015 (4.94 t ha⁻¹) and 2016 (5.38 t ha⁻¹), and it was significantly lowest in 2017 (3.87 t ha⁻¹). Straw yield was highest in 2016 (2.86 t ha⁻¹), and it exceeded the values noted in the remaining years by 16%. The harvest index was similar in 2015 and 2016, and it was 9% lower in 2017. Grain yield increased by 30% and 36%, whereas straw yield increased by 20% and 35% in response to the nitrogen rates of 80 and 120 kg ha⁻¹, respectively. A minor increase in grain yield (3%) was observed in treatments with a sowing density of 550 seeds m⁻² relative to the remaining sowing densities.

Path modelling

A simple correlation analysis of manifest variables in all phenological stages revealed significant correlations between the LAI and leaf greenness (SPAD) only in stage Z32, as well as a very strong correlation between the net photosynthetic rate and the transpiration rate, which was positive in stages Z32 and Z45 and negative in stage Z59. No simple correlations were noted between the indicators of physiological processes (Pn, E, WUE) and biophysical parameters (LAI, SPAD). AAll correlations between the manifest variables of yield components and biological yield were statistically significant, excluding the correlation between stem length and ear length (Supplementary information).All correlations between the manifest variables of yield components and biological yield were statistically significant, excluding the correlation between stem length and ear length (Supplementary Information). The outer and inner PLS-PM models well fit the data, and their goodness of fit was determined at 0.973 and 0.786, respectively. The outer weights provide information about the relative importance of a manifest variable for the corresponding latent variable (for details please see the Supplementary Information). Outer weights that exceed 0.3 are considered meaningful. By the same token, loading estimates represent the correlations between a latent variable and the corresponding manifest variables. Loadings higher than 0.7 capture more than 50% of the variability contributed by a latent variable to the corresponding manifest variable. In general, both indicators in the outer model, i.e. outer weights and loadings, exceeded the thresholds, which indicates that manifest variables were strongly related with latent variables. Growth regulators (({w}_{GR}) = − 0.007) and the length of the growing season (({w}_{DAYS})=0.197) provided the only evidence for the low explanatory value of latent variable A (agronomic factors).

In the inner model, all equations that regressed latent variables well fit the data and were statistically significant (Table 3). The latent variables expressed by the value of R² increased in successive stages of T. durum growth and development, from 0.218 in physiological processes in stage Z32 (Table 3, Eq. 1) to 0.698 and 0.708 in yield components and Biological Yield, respectively (Table 3, Eqs. 7 and 8). It is worth noting that in successive stages of growth, the value of physiological processes was relatively lower in comparison with biophysical parameters.

The analysis of path coefficients (β_i) revealed that agronomic factors (A) and climate conditions (CC) in stages Z32, Z45 and Z59 exerted a specific influence on physiological processes (PP) and biophysical parameters (BP) of T. durum plants. Agronomic factors directly determined physiological processes in all stages and biophysical parameters in stages Z32 and Z59. At the same time, climate conditions did not exert a direct influence on physiological processes in any stage, but directly affected biophysical parameters in all stages. All of the modeled parameters, i.e. agronomic factors, climate conditions and physiological processes, significantly influenced biophysical parameters in stages Z32 and Z59, but not Z45. Consequently, it can be stated that agronomic factors were the main determinant of variability in physiological processes (photosynthesis, transpiration) in a model evaluating the impact of agricultural practices on yield and the manifest variables associated with T. durum growth and development. At the same time, physiological processes made a significant but negative contribution to biophysical parameters. A one unit increase in photosynthesis processes with constant values of agronomic factors and climate conditions implies a decrease of − 0.382, − 0.065 and − 0.395 in biophysical parameters in stages Z32, Z45 and Z59, respectively.

The performance of every preceding latent variable in terms of its total impact on the target latent variable, i.e. the biological yield of T. durum (IPMA – Importance-Performance Map Analysis), was analyzed to highlight latent variables associated with agricultural practices that improve biological yield. The total effect (importance) of preceding latent variables (A, CC32, PP32, BP32, CC45, PP45, BP45, CC59, PP59, BP59, YC and CC) on the anticipated performance of the specific target (Biological Yield) is presented in Fig. 3.

The importance and performance of latent variables that influenced the biological yield of T. durum varied. The biological yield of T. durum was affected mostly by agronomic factors (A), followed by yield components (YC) and biophysical parameters (BP) in growth stages Z59 (BP59) and Z32 (BP32), climate conditions in stage Z59 (CC59), and climate conditions in stage Z32 (CC32). A one unit increase in the above latent variables led to an increase of 0.575, 0.422, 0.234, 0.203 and 0.109 units in biological yield, respectively. At the same time, the performance scores of these latent variables were determined at 53.1, 53.5, 67.1, 59.6 and 61.8, respectively (scores closer to 100 denote higher performance). The remaining latent variables, in particular climate conditions for the entire growing season and physiological processes in stage Z32, were characterized by low importance and exerted a relatively small effect on biological yield performance.

The results of the importance-performance analysis clearly indicate that latent variables have considerable potential to optimize the agricultural conditions for the growth and development of T. durum plants.

Source: Ecology - nature.com