An exploratory method to determine the plant characteristics affecting the � nal yield : A study on Echium amoenum Fisch . & C . A . Mey . under different fertilizers application and plant densities


 Background: Employing of advanced statistical methods to quantify agricultural information has helped to carry out targeted planning to alleviate the problems of farmers, researchers and policy section. One of these exploratory methods, is multivariate statistical analysis that examines and models the relationship between variables. Considering the importance of Echium amoenum and its use growing trend in traditional medicine and the pharmaceutical industry, also the lack of information on the correlations between its yield and morpho physiological traits, the objective of this study was to determine the causality path in which the Echium amoenum characteristics affects the yield of Echium amoenum as regards of application of organic and chemical fertilizers under different plant densities.Results: The employed method revealed that organic fertilizers increased flower yield compared with the control. The flower yield as a result of application of compost, vermicompost and cattle manure were increased by 25, 28, and 27 percent compared with the control, respectively. The results of multiple regression showed that variables of plant height, shoot dry weight, flower number per plant were the main factors affected the flower yield. The relative contribution of shoot dry weight was 16 and 25 percent more than plant height and flower number per plant, respectively. Conclusions: Causality analysis identified that shoot dry weight per plant had indirect effect on flower yield in different paths, as mainly was imposed through plant height considering the path coefficients. this study suggests that optimum production of Echium amoenum with application of ecological inputs along with effective agronomical managements of the causal paths of flower yield forming, including increase in shoot yield and plant height could be achieved through an ecological cropping system with reduced costs and no health concerning due to agrochemicals residual.


Introduction
Currently, the use of eco-friendly inputs as approaches to achieve sustainable agriculture has been considered (Wu et al., 2005;Altieri et al., 2017). Undoubtedly, the application of organic fertilizers and manures especially in nutrient-poor soils in addition to its positive effects on all soil properties and the increase of its organic matter, can also be bene cial in economic, environmental and social aspects and can be a suitable and desirable substitute for chemical fertilizers in long-term (Mao et al., 2008;Lee, 2010). Application of animal manure increases soil yield, improves water holding capacity and improves soil aggregation and increases water use e ciency and increases crop yield (Raja Sekar and Karmegam, 2010). Organic compost and vermicompost in most parts of the world have been successfully used on a large number of agricultural products (Shamsodin et al., 2007;Doan et al., 2013). It was reported that the highest root growth rate and root relative growth rate of sorghum (Sorghum bicolor L.) were resulted from the combined application of vermicompost and mycorrhiza treatment (Kamaei et al., 2019). Introducing these organic fertilizers to soil improves nutritional, physical, chemical and biological aspects of soil ecosystem are also improved (Robin et al., 2001;Singh et al., 2011;Coleman and Callaham, 2017). In a study the effect of different organic fertilizers on the quantitative and qualitative properties of several medicinal plants was investigated, it was reported that vermicompost increased Echinaceae purpurea L. plant height and increased fresh and dry shoot weight of Melissa o cinalis L. (Delate, 2000). It was reported the application of organic fertilizers, especially vermicompost, on Calendula o cinalis L. resulted in a signi cant increase in the number of branches per plant and the ower number per plant (Rezaee and Baradaran, 2011). It was reported that application of vermicompost increased dry matter of tomato (Lycopersicum esculentum L.) also improved nitrogen uptake e ciency in organic tomato production (Ebrahimi et al., 2019).
Plant density, as an agronomical management practice, plays an important role in the yield of different crops, so identifying the optimum plant density is one of the basic principles of crop production (Ibrahim, 2012). A balanced increase in plant density will accelerate canopy closure, increase leaf area, increase productivity of environmental resources, reduce weeds and ultimately improve the yield and yield components of different plants (Ndabamenye et al., 2013;Chauhan and Abugho, 2013). It is remarkable that at high plant densities, leaf loss rate increased and consequently. It has negative effects on the quantitative and qualitative characteristics of the plant, due to shading and competition of plants for light and scarcity of available resources, as well as greater susceptibility of plants to pathogens, (Zhang et al., 2012). In a study, the effect of different plant densities (12.5, 16.6 and 25 plants m − 2 ) on yield and yield components of Coriander (Coriandrum sativum L.) was studied and reported that, the number of umbrellas per plant, 1000-seed weight and plant dry weight decreased with increasing density (Akhani et al., 2012).
In another experiment, the effect of different plant densities on yield and yield components of a medicinal plant (Hibiscus sabdariffa L.) was studied and reported that with increasing the distance between planting rows from 50 to 100 cm, ower yield increased (Mir et al., 2011).
In recent years, the use of advanced statistical methods to quantify agricultural information has helped to carry out targeted planning to alleviate the problems of farmers, researchers and policy section. One of these methods, which is widely used in all sciences disciplines today, particularly agricultural sciences, is multivariate statistical analysis that examines and models the relationship between variables (Kerlinger and Pedhazur, 1973). One of the exploratory multivariate methods as an elaborating statistical method in analyzing and explaining many phenomena is causality analysis or path analysis method. Causality analysis is a precise analytical tool to determine the share of direct and indirect effects of one variable with the other ones, since many studies have found that one variable not only has a direct effect on the other variable, but also indirectly through other variables affects those (Toebe and Filho, 2013;Guler et al., 2001). In other words, causality analysis divides the correlation coe cient between the two variables into a direct and indirect effects (Cramer, 2003;Everitt and Dunn, 1991). In a study, the relationships between yield and yield components of spring sa ower (Carthamus tinctorius L.) were investigated and reported that biological yield, number of pod and branch and number of grains per pod were effective on grain yield, also according to the results of stepwise regression, whereas path analysis results showed that only two of the four traits (biological yield and number of pods per plant) effectively affected grain yield (Omidi Tabrizi, 2003).
Echium amoenum Fisch. & Mey. is a perennial herb, Boraginaceae family plant, and is a valuable herb due to excellent medicinal properties (Mehrabani et al., 2005). This plant has been distributed across the northern parts of the country as wild vegetation (Sayyah et al., 2009). In traditional medicine, the petals of this plant are used as diuretics, analgesics, diaphoretic, and treat for high blood pressure (Nooriyan Soroor et al., 2013;Hornok, 1992). Considering the importance of Echium amoenum and its use growing trend in traditional medicine and the pharmaceutical industry, also the lack of information on the correlations between its yield and morpho physiological traits, this study was conducted aimed to determine the causality path in which the borage characteristics affects the yield of Echium amoenum as regards of application of organic and chemical fertilizers under different plant densities.

Site description
Field studies were conducted during the 2013-14, 2014-15 and 2015-16 growing seasons at the Research Farm Station of Agriculture Faculty, Ferdowsi University of Mashhad, Iran (latitude: 36° 15′ N; longitude: 59° 28′ E; elevation: 985 m above sea level). The Research station was located in Kashaf-rood watershed in northeast of the country in a semi-arid region with mean annual precipitation of 252 mm and temperature of 15° C. Documented declaration of cropping history of the land which experiment was conducted in con rmed that it had been under fallow for the past three years, with no agrochemicals chemicals consumed or imported in (Research Station Archive).
Soil samples were taken at 0-30 cm depths and analyzed for some physiochemical properties (FAO, 2008) before conducting the experiment (Table 1).  Minimum tillage was carried out to prepare the soil with emphasis on ecological soil cultivation operations, so that after a shallow disk, plots of 2.5×5 m with a distance of 1 meter between, to avoid nutrients mixing due to irrigation consisting of 6 rows were arranged to sow the borage seeds on the middle of rows.
To applying organic fertilizers, the amounts of NPK in compost, vermicompost and cattle manure were determined (the results of the analysis of organic fertilizers used in the experiment shown in Table 2), then according to NPK requirements of Echium amoenum (Najafpoor Navaee, 2002) as well as taking into account the local farmers recommendations, the needed amounts of fertilizers were determined. Pure nitrogen by 90 kg ha − 1 (this amount of pure nitrogen was provided by 180kg urea fertilizer containing 46% N), half of which at the time of sowing and the other half after thinning operation were applied, while in the second cropping year (2014-15), the same amount of fertilizer was added in two stages (beginning of regrowth and four-leaves stages in the second year).
In late February 2013, organic fertilizers were broadcasted on the soil surface uniformly and immediately were mixed into the soil (a depth of 30 cm) of related plots using a spade. In late February 2014, to promote the plants regrowth, the same amount of fertilizers were added into the soil on the side of planting rows of the related plots in a depth of 15 cm. At the spring of the second and third years (2015-16) during of the owering season (April 6, to June 20) the owers of all experimental plots were harvested daily (borage is identi ed with an undetermined growth pattern) then fresh and dry weights of owers were measured. Harvested owers were air dried under the shadow avoiding direct sunlight. The total dry weight of owers during the owering period was considered as the dry ower yield per plot. Three plants per plot were randomly selected and the ower number were counted during the owering period.
At the end of the growing season, with the onset of seed ripening and plant shoot drying, three plants were randomly selected from each plot and traits including shoot yield, branch number, branch length, plant height and their canopy diameters were measured. To determine seed yield, total plants of all experimental plots were harvested and seed weight was determined. Data statistical analysis A normality test was already performed. Transformation was also performed for numerical data where needed. To ensure uniformity of treatment variances, the Bartlett's test was performed. Since there was no statistical difference between experiment data of two years (2014-15, 2015-16), thus the mean of each trait values during two years were reported. Analysis of variance (ANOVA) and graph plotting were done using SAS Ver.9.1, Slide Write Ver.2 and Microsoft Excel Ver. 14. All mean comparisons were performed by Duncan's multiple range test (DMRT) at 5% probability level. Growth characteristics affecting dry ower yield were determined using multiple regression and Minitab Ver.16 software. In order to nd out the causal relationships between yield and growth characteristics affecting it, causality analysis was performed (Everitt and Dunn, 1991).

Shoot yield
The effect of plant density on shoot yield was signi cant (Table 3), as the highest shoot yield per plant resulted from medium density (5 plant m − 2 ). This plant density increased yield by 34 and 47%, compared with densities of 10 and 3 plant m − 2 , respectively. Different organic and chemical fertilizers had a signi cant effect on shoot yield ( Table 3). All the organic fertilizers had a positive effect on shoot yield as shoot yield as a result of application of compost, vermicompost and cattle manure increased by 25, 7 and 19%, respectively, compared with the control. The chemical fertilizer also resulted in 17% increase in shoot yield compared with the control. Compost and cattle manure increased shoot yield by 10% and 2%, respectively compared with chemical fertilizer.  In each column, means followed by the same letters are not signi cantly different (p ≤ 0.05), at 5% probability level based on Duncan's multiple range test.
With decreasing of density, the trend of shoot yield changes in different organic fertilizers was similar, so that under all organic fertilizers application condition, by decreasing densities down to 5 and 3 plants m − 2 , shoots yield was increased rst and then decreased ( Table 4). As it shown in Table 4, chemical fertilizer in 5 and 3 plant m − 2 increased shoot yield by 22 and 62%, respectively compared with the control.

Plant height
Although the effect of different densities on plant height was not signi cant, plant height was affected by different organic and chemical fertilizers ( Table 3), so that all organic fertilizers increased plant height compared with the control. Application of compost, vermicompost and cattle manure resulted in increased plant height by 21, 12 and 17%, respectively, while the average plant height under these fertilizer applications was more than the control by 12%. All of the organic fertilizers had superiority to chemical fertilizer regarding plant height, so the plant height as a result of application of compost, vermicompost and cattle manure were higher than chemical fertilizer by 10, 4 and 6%, respectively.
As it shown in There was a signi cant difference between different organic and chemical fertilizers as regards the effect on ower number per plant (Table 3), as vermicompost and cattle manure increased this trait by 19 and 29%, respectively, compared with the control. Application of these organic fertilizers also resulted in increased ower number per plant compared with the chemical fertilizer.  (Table 4). Table 4, fertilizer in all densities had no signi cant effect on ower number per plant compared with control, but it seems that at 5 plant m − 2 had effective impact on plant density as its application in 5 plants m − 2 resulted in increased ower number per plant by 3 and 36%, respectively, compared with application of fertilizer in 10 and 3 plant densities, respectively.

Dry ower yield
Dry ower yield was signi cantly affected by plant density (Table 3), as the highest ower yield (3972.6 kg ha − 1 ) was obtained from density of 10 plants m − 2 which was higher than densities of 5 and 3 plants m − 2 by 21 and 47%, respectively. The effect of organic and chemical fertilizers on dry ower yield was signi cant (Table 3), as all organic fertilizers increased dry ower yield compared with control. Dry ower yield resulted from application of compost, vermicompost and cattle manure were higher by 25, 28 and 27% than the control, respectively. It is remarkable that all organic fertilizers (compost, vermicompost and cattle manure) had higher dry ower yield than chemical fertilizer by 24, 27 and 26%, respectively.
The interaction of plant density and fertilizers on dry ower yield was signi cant (Table 3), as by decreasing plant density, the e ciency of organic fertilizers in increasing dry ower yield was increased. Organic fertilizers had no signi cant effect on dry ower yield in density of 10 plants m − 2 , whereas compost, vermicompost and cattle manure increased dry ower yields by 41, 49 and 40%, respectively compared with the control. In density of 3 plants m − 2 application of compost, vermicompost and cattle manure to soil increased dry ower yield by 53, 44 and 69%, respectively compared with the control ( Table 4). Table 4, the effect of organic fertilizers was different at different among plant densities, as in densities of 10 and 5 plants m − 2 application of vermicompost, and in density of 3 plants m − 2 cattle manure application resulted in more dry ower yield than other treatments.

Relative comparison of growth characteristics of borage under plant densities
Relative values of growth characteristics of borage under different densities shown in Fig. 1.
The advantage of the density 5 plant per m − 2 considering fresh ower yield, shoot dry yield, dry ower yield is clearly revealed.
Coe cients of correlation between traits As shown in Table 5, the number of branches per plant was the only trait that had a signi cant positive correlation with the ower number per plant. Correlations between fresh and dry weight of owers per plant with all studied traits were signi cant except of ower number per plant and canopy diameter. Seed yield was also correlated with most physio morphological traits. The shoot yield per plant was signi cantly correlated with fresh and dry ower weight per plant, seed yield and branch length ( Table 5).
Although there was a signi cant correlation between the number of branches per plant and most of the studied traits, its correlation with fresh weight (r = 0.55**) and ower dry weight per plant (r = 0.56**) was more than the correlations of this trait with the other ones. Plant height and branch length were correlated with most of the studied traits, but the canopy diameter was not signi cantly correlated with any of the studied traits ( Table 5).
The correlation between most studied morphological traits and dry ower yield was positive and signi cant, as the highest correlation was related to shoot yield per plant (r = 0.42**), ower fresh weight (r = 0.40**), and ower dry weight per plant (r = 0.36*), and plant height (r = 0.35*) ( Table 5). The correlations of number of branches per plant, branch length and ower number per plant were also signi cant with dry ower yield, so by increasing each of them, dry ower yield would be improved (Table 5). Identi ed growth characteristics affecting dry ower yield using multiple regression The results presented in Table 5 showed that the ower yield of Echium amoenum was correlated with most of the measured variables. Accordingly, multiple regression was used to analyze the relationship between ower yield as a function variable (Y) and traits affecting it (independent variables, X). After identifying the main growth characteristics affecting dry ower yield using multiple regression, the direct and indirect effects of each of these characteristics were estimated using causality (path) analysis method. The effect of these characteristics on each other and on dry ower yield has been shown in Fig. 2.
Equations no. 2 to 7 were used to calculate the coe cients of direct effects of growth characteristics on each other and on dry ower yield shown in Fig. 2  Calculated values of the direct and indirect effects of each of growth characteristics and analyzed correlation coe cients between these traits and dry ower yield are presented in Table 6.  (Fig. 3).
As it shown in Fig. 4, there was a linear function between shoot weight per plant and plant height (Fig. 4a). On the other hand, increased plant height led to improved dry ower yield (Fig. 4b), so it is reasonable to expect that dry ower weight increases with increasing shoot yield.

Discussion
It seems that by reducing the number of plants per square meter, the positive effect of organic fertilizers on the growth characteristics of the plant was revealed obviously, probably due to more availability and synchrony of the nutrients in organic fertilizers to the plant needs. It seems that with decreasing plant density, plant access to growth resources such as light, water and nutrients is increased and resulted to improved growth characteristics of the plant including its height. Some researchers studied the effects of organic and chemical fertilizers on yield and essential oil percentage of basil (Ocimum basilicum L.) and reported that vermicompost-treated plants had higher plant height, leaf yield, shoot yield, fresh and dry yield than other treatments ( It seems that the average plant density played the most role in increasing the ower number per plant. In low plant density, ower number per plant decreased compared with the average plant density probably due to excessive access to food and growth resources by the plant. In high plant density, this decrease was related to plant competition over water and nutrients and lack of e cient use of resources. Organic fertilizers appear to be likely to increase the ower number per plant by supplying the plant with the micro nutrients ( , which eventually led to increased owering. The effects of different organic and biological fertilizers on the sa ower were studied and it was reported that vermicompost solely or combined with Nitroxin® and Nitrajin® biofertilizers improved the quality and quantity of the plant (Rezvani Moghaddam et al., 2013). In a same study, it was reported that application of 10 t ha − 1 vermicompost increased ower number, plant height, 1000-seed weight, biological yield and essential oil content of Foeniculum vulgare Mill. (Darzi et al., 2006).
In high plant densities, it seems that intra-speci c competition was increased and growth resources, particularly radiation, would not been adequately provided to the plant (Ndabamenye et al., 2013), thus resulted in a decrease in dry ower yield. In a study, the effect of distance between planting rows (60, 70 and 80 cm) and within rows (25, 35 and 45 cm) on yield and yield components of Satureja khuzistanica Jamzad, was investigated. The results showed that the highest ower yield and canopy diameter were observed in 45 cm within row distance and density of 7 plant m − 2 had the highest dry matter yield (Hekmati et al., 2012).
The vermicompost probably played an important role in supplying the water needed for the plant (Shamsodin et al., 2007) because of its high moisture holding capacity, thereby producing more ower dry yield. Cattle manure at low levels of plant density signi cantly increased dry ower yield probably through increased nitrogen release in soil (Motta and Maggiore, 2013). Some studies have shown that the application of organic fertilizers reduce the salinity effects and increase the uptake of phosphorus and nitrogen thus improve the qualitative and quantitative characteristics of plants (Sabahi et  Flower yield is a complex feature that is in uenced by many physiological processes and its measurable performance would be revealed in phenological, morphological, and physiological traits (Hobbs and Mahon, 1982). Weak correlations between some traits appeared to be related to differences in the time of traits measured, as traits such as ower number per plant and ower weight per plant were measured during owering, while the traits such as plant height, number of branches per plant and canopy diameter were evaluated at the end of the owering period. Therefore, causality analysis was performed to accurately determine the contribution of each of the traits to improvement of dry ower yield.
Although the yield of most crops, particularly medicinal plants, has increased over the past decades, but the morphological and physiological processes underlying this increase of yield are not well identi ed (Tollenaar, 1991). Researches revealed the positively correlation between the physio morphological traits and yield of medicinal plants including Mentha pulegium, Peppermint (Mentha piperita) and Thymus vulgaris (Mirzaee Nadooshan et al., 2001;Kukreja et al., 1992). If the origins of increased yield of medicinal plants are identi ed, paths to improve their actual potential by better crop management practice and effective nutrition supply may be identi ed (Fraser and Eaton, 1983). In this study physio morphological traits affecting yield Echium amoenum were identi ed using multiple regression and causality analysis.
The coe cients of Eq. 1 show the relative impact of changes in each of the variables in the model on ower yield. For example, the change in ower yield was 0.32 units per unit of change in shoot yield per plant, while this change would be 0.27 per unit increase in plant height. In other words, the relative share of shoot yield per plant was about 16% higher than that of plant height, implying some important growth characteristics of borage such as producing numerous branches and ower formation at the end of them and nally the effect of these traits on ower yield. However, to better interpret these results, the unit of measurement for each variable should also be considered, which is why multiple regression was performed on standardized traits data. Thus, due to the effects of different treatments (different types of fertilizer and plant densities) in the above model, it is possible to quantitatively evaluate the response of borage based on the rate of increase or decrease of the variables affected by the treatments.
According to Table 6, direct effect of shoot yield per plant on dry ower yield was more than direct effect of plant height via ower number per plant. Plant height had more direct effect on dry ower yield than ower number per plant.
Shoot weight per plant affected dry ower yield indirectly in three ways: (A) Indirect effect of shoot weight per plant through plant height (P 31 + P 42 ) (B) Indirect effect of shoot weight per plant through ower number per plant (P 31 + P 43 ) (C) Indirect effect of shoot weight per plant by plant height and number by owers per plant (P 21 + P 32 + P 43 ) As it shown in Table 6, shoot weight per plant had the most indirect effect on dry ower yield (0.0594), which resulted in increased dry ower yield mediated through plant height. It was reported that the yield of owering branches of Camphorosma monspeliaca L. was positively correlated and affected by shoot yield. Also, the number of tillers that had the most direct effect on the yield of owering branches was also indirectly affected by plant height (Abbaszadeh et al., 2011).
The results (Table 6)  Comparison of morphological traits affecting dry ower yield showed that shoot weight per plant affected dry ower yield more than the other traits, as the total direct and indirect effects of shoot yield per plant was more than the other traits. Considering the paths coe cient affecting dry ower yield (Fig. 2), it seems that management practices and treatments that would increase shoot yield per plant would lead to improved yield of Echium amoenum. Results of a study on sun ower (Helianthus annus L.) showed that there were positive correlations between biological yield, shoot yield and grain yield (Amirian et al., 2013). In another study, shoot yield, plant height and number of grains per plant were identi ed as the most in uential traits on the yield of Trigonella foenum-graecum L. (Singh et al., 2012). Some researchers reported that shoot yield, particularly umbrellas dry weight and 1000-seed weight of Coriandrum sativum L. were the most important traits affecting the yield of this medicinal plant (Dyulgerov and Dyulgerova, 2013). Guler et al., (2001) determined negative and signi cant relationship between 100-seed weight and seed yield of chickpea using path coe cient analysis.

Conclusions
The results showed that cultivation of borage with density of 5 plants m − 2 and application of compost resulted to the highest ower yield. There was a signi cant positive correlation between dry ower yield and all studied growth characteristics except seed yield and canopy diameter. According to the results of multiple regression, shoot yield per plant, plant height and ower number per plant were identi ed as the main factors affecting dry ower yield, although the relative proportion of plant height compared with shoot weight per plant and ower number was higher by 16 and 25 percent, respectively. Causality analysis revealed that shoot weight per plant had the most direct effect on dry ower yield, while this trait through three paths (1-plant height, 2-ower number per plant, 3-Plant height, and then ower number per plant) had an indirect effect on dry ower yield. The causality analysis also identi ed that shoot weight per plant seems affected dry ower yield through plant height, along with increasing shoot weight per plant, plant height was increased which in turn improved dry ower yield.
Conclusively, this study suggests that optimum production of Echium amoenum with application of ecological inputs along with effective agronomical managements of the causal paths of ower yield forming, including increase in shoot yield and plant height could be achieved through an ecological cropping system. Moreover, achieving more yield from organic fertilizers application than chemical fertilizer in this study, promises agrochemicals free and healthy production of this medicinal plant could be achieved from low input cropping systems or marginal farms using ecological inputs.   Path coe cients diagram showing the causality path of dry ower yield forming and growth characteristics of Echium amoenum as a result of application of organic and chemical fertilizers and different plant densities (e represents non-measurable errors).