Variability in soil properties influencing pigeonpea (Cajanus cajana L.) yield: a multivariate statistical analysis

Study setting

The study area was located in the Northern dry zone of Karnataka State, India (Figure 1), between latitudes and longitudes of 17°40’17.956’’ N - 76°59’46.836” E and 17°38’12.689” N - 77°1’17.049” E, covering 645.20 ha, at an altitude of 511 - 637.86 m above mean sea level. The study was planned for two site years viz., during kharif (cropping season between June to October) 2018–2019 and kharif 2019–2020, to assess the variability in surface soil properties for each soil phase and crop growth parameters influencing pigeonpea yield.

Figure 1. Location map with soil physiography interpreted imagery of LISS IV merged Cartosat-1.

Kalmandari Tanda-1 micro-watershed (MWS) reported mean monthly minimum and mean monthly maximum temperatures of 19.08°C and 31.97°C, respectively, during January 2018 and May 2018, whereas the reported values for these variables were respectively 18.89°C and 33.7°C for January 2019 and May 2019. The average annual rainfall measured by Kalmandari Tanda-1 MWS was 442 mm and 831.50 mm in 2018 and 2019, respectively (Figure 2 and Figure 3).

Figure 2. Mean monthly weather parameters of Kalmandari Tanda-1 MWS during 2018.

Figure 3. Mean monthly weather parameters of Kalmandari Tanda-1 MWS during 2019.

Experimental procedures

A comprehensive methodological workflow adopted in this study is presented in Figure 4. To divide the entire micro-watershed into different soil phase units, a detailed land resource inventory was initially carried out at a 1:8000 scale, using Indian Remote Sensing Satellite-P6 (IRS P6) Cartosat-1 merged LISS-IV satellite imagery (2.5 m spatial resolution) as base map to interpret the soil physiographic unit map (Figure 1). The soil (soil profile depth, number of horizons, soil color, soil structure, texture, consistency, presence of carbonates, and soil pH) and site (slope, erosion, drainage, runoff, gravelliness, stoniness, presence of rocks on surface, lithology/parent material and current land use) characteristics were recorded for all soil profile sites on a standard pro forma (Pedon description form which consists of a list of soil and site parameters that need to be recorded during a field study. A copy is provided in the Data availability section (see Extended data, Rajesh, 2021)) as per the guidelines given in Field Guide for Land Resources Inventory (Natarajan et al., 2016).

Figure 4. Methodological workflow.

In this micro-watershed, both Vertisol and Alfisol soils were found, which are soils derived from basalt and basalt-laterite intrusion parent materials, respectively (Government of India, 2008). Pedons that showed similar soil profile horizons in soil color, soil texture, soil pH, and consistence, mineral and chemical compositions were grouped into soil series (FAO, 2006) and further, these soil series were divided into 23 soil phases (Figure 5), based on surface characteristics with respect to soil texture, slope, erosion and gravelliness.

Figure 5. Soil phases of Kalmandari Tanda -1 micro watershed.

A total of 462 surface soil samples (0–30 cm) were collected in two site years. Each year, before sowing of pigeon pea, 231 soil samples were collected from the same locations (recorded using Trimble Juno 3D, a hand-held GPS) of each soil phase of the micro-watershed. Physico-chemical properties of all collected surface soil samples were analyzed using standard procedures. Erosion hazards were judged through the visible soil erosion method, by assessing the presence of rills and gullies within a field, as well as their associated deposits (Evans, 2013). Soil texture was evaluated using the feel method, and slope with the help of a dumpy level. Organic carbon (OC) was determined using the Walkley & Black (1934) wet oxidation method. Available N was determined by a modified alkaline potassium permanganate method as described by Subbiah & Asija (1956). Available P₂O₅ was determined using Olsen’s method. Available K₂O was estimated using a flame photometer after extraction with ammonium acetate. Soil reaction (pH) was determined in 1:2.5 soil water suspensions using a glass electrode (Piper, 1966). Electrical conductivity was measured in the soil water (1:2.5) suspension using a conductivity bridge (Jackson, 1973). The level of free calcium carbonate ions in soil samples was determined using a rapid titration method with standard HCl (Piper, 1966).

Cationic micronutrients like iron, copper, manganese and zinc were extracted using diethylene triamine penta acetic acid (DTPA, 0.005 M and 0.01 M CaCl₂ + 0.1N tri-ethanol-amine at pH 7.3), and the concentration was measured using an atomic absorption spectrophotometer (ContrAA 700 Make) as outlined by Lindsay & Norvell (1978). Available boron in soil was estimated with a colorimetric method using hot water as extractant and expressed in mg kg^-1 (Page et al., 1982). Available sulphur in soil was extracted with CaCl₂.2H₂O (0.15%), and the extract was reacted with barium chloride crystals. The intensity of the resulting turbidity was measured using a spectrophotometer at a wavelength of 420 nm (Jackson, 1973)

The bulk density was estimated using Keen’s cup method for the disturbed soils (Piper, 1966). Particle size distribution of soil samples was determined using the "International Pipette" method (Piper, 1966); the soil’s cation exchange capacity (CEC) was determined by equilibrating the soil with 1N sodium acetate solution using a flame photometer, to calculate the CEC. The exchangeable sodium percentage (ESP) was calculated by dividing the exchangeable sodium by CEC and exchangeable bases (Ca, Mg, and K, Na), which were measured by titration and flame photometer respectively.

Pigeon pea growth and yield parameters were recorded using a crop cutting experiment, where observations were regularly made in the farmer’s field at specific locations of the study area and at different crop growth stages. The growth parameters viz., plant height (in centimeter), number of branches per plant, and leaf area index (AccuPAR Ceptometer model LP-80) were recorded for each soil phase. Five pigeon pea plants were labeled in a 1 m² area of each soil phase unit and the growth parameters were recorded at 30, 60, and 90 days after sowing (DAS), as well as during the crop harvesting stage. Similarly, pigeon pea yield parameters, namely the number of pods per plant and grain yield (kg ha^-1), were recorded at the harvest stage and after the harvest respectively. Details of pigeon pea varieties and fertilizer doses (urea and di-ammonium phosphate manufactured by Zuari Agro Chemical Ltd., India) used are given in the Extended data (Rajesh, 2021). Meanwhile, soil moisture and available water content (AWC) at each soil-phase were measured using a theta probe (Stevens Water Monitoring Systems Inc, S/N 238825) and pressure plate membrane apparatus (SOILMOISTURE Equipment Corp, Model #1500F2), respectively.

Statistical analysis

Descriptive statistics and principal component regression analysis of soil and plant parameters were carried out in SPSS v.16. An assessment of the normality of the data is pre-requisite for PCA. A Z-scale test was applied for normality test using skewness and kurtosis (for n < 300). The distribution is approximately normal if skewness and kurtosis are between -1 and 1. The right and left skewed data (Table 1 and Table 2) were subjected to logarithmic and square root transformation to improve the normality. These transformed variables were subjected to a PCA with varimax rotation (Ayoubi et al., 2009; Cox et al., 2003; Shukla et al., 2004b). The descriptive statistics of physico-chemical properties of soil and plant growth parameters are presented in Table 1 and Table 2, respectively.

Multivariate analysis

PCA was used as a data reduction technique. The independent variables such as soil properties (soil pH, EC, CaCO₃, OC, available N, P₂O₅, K₂O, S, Zn, Fe, Cu, Mn, B, exchangeable Ca, Mg, Na, ESP, CEC [Cmol(p⁺)kg^-1], Sand [%], Silt [%], Clay [%], bulk density [Mgm^-3], permanent wilting point [PWP in %], field capacity [FC in %], AWC [%]) and plant growth parameters (plant height, number of branches per plant, LAI, number of pods at harvest stage) were subjected to a PCA; variables with factor loadings greater than the measured Kaiser-Meyer-Olkin (KMO) values, were chosen as highly correlated variables from the derived principal components (PCs) with eigenvalues greater than one (Suryanarayana & Mistry, 2016). The KMO test is a measure of sampling adequacy and factorability. The KMO tests whether the partial correlations within the data are close enough to zero to suggest that there is at least one latent underlying factor among variables (Vijayamohanan Pillai & Asalatha, 2020). As a KMO value of < 0.5 is unacceptable (Hair et al., 2003), the results of the PCA with a KMO > 0.5 loadings indicate that the chosen correlation matrix was appropriate for factor analysis.

The Kaiser Meyer Olkin (KMO) test is given by;

Equation 1: $M{O}_j=\frac{{\sum }_{i\ne j}{r}_{ij}^2}{{\sum }_{i\ne j}{r}_{ij}^2+{\sum }_{i\ne j}U}$

Where, r = [r_ij] - Correlation matrix

            U = [u_ij] - Partial covariance matrix

The eigenvalue is the variance explained by a component or factor and is denoted by lambda.

Equation 2: $\lambda \kappa ={\displaystyle \sum _{i=1}^n{{\rm A}}^{2}i}k$

Where, A²ik is the factor loading for variable i on component k, and n is the number of variables. A low eigenvalue represents a lower contribution to the variance in the analyzed set of variables. The component matrix table shows the component loadings, which represent the correlations between the variables and the components. As a rule of thumb, the larger the size of the component loading for a variable, the more significant the variable is in the interpretation of the component (Nargundkar, 2005). The first component is generally more highly correlated with the variables than the second components, and so on. A varimax rotation enhances the interpretability of the uncorrelated components; therefore, only the outcomes of the rotated component matrix are presented in our results.

Linear principal component regression (LPCR) analysis

The derived PCs were subjected to a stepwise linear regression analysis, which is an alternative to multiple regressions analysis. In LPCR, instead of regressing the dependent variables on the explanatory variables directly, the principal components of the explanatory variables were used as regressors, each time removing the weakest correlated variable. The highly correlated and significant soil and plant independent/explanatory variables from each PC were fitted to the dependent variable (pigeon pea crop yield). Further linear equations were drawn separately.

PCA & LPCR to assess the influence of soil properties on pigeon pea yield

Seven components were extracted with total cumulative variances of 82.30% and 80.75% during 2018 (Table 3) and 2019 (Table 4), respectively. The principal component 1 (PC1) in each site year was identified as soil available nutrient content with high loadings of K₂O (0.843 and 0.764, respectively), Mn (0.870 and 0.882 respectively), Ca (0.845 and 0.900), and Mg (0.835 and 0.919 respectively). The presence of these adsorbed cations on negatively charged clay surfaces (supported by high loadings of clay content (0.905and 0.929) contributed to significantly high CEC loadings (0.897 and 0.937) (Schoonheydt et al., 2018). High clay content supported high available water content with loadings of 0.822 and 0.625 during 2018-19 and 2019-20 respectively. Therefore, sand (-0.858 and -901) and bulk density (-0.927 and -0.743) are negatively correlated in PC1 of both the site years. Though the initial mean available water content is high (27.93 %) in soils during 2018–19 (Table 1), the scarcer rains in subsequent months during 2018–19 (Figure 2) has inferred that there was no sufficient soil moisture to dissolve the soil nutrients to make it readily available to pigeonpea plant when compared to the rains of 2019–20 (Figure 3). In 2018, PC2 was identified as humification and vegetative growth with high loadings of OC and available N. The high loading of EC during 2018 may be due to the retention of dissolved salts associated with low precipitation. In 2019, because of higher precipitation levels (Figure 3) the results of PC2 converged towards enhanced calcification with high loadings of CaCO₃ (0.793) due to dissolution of CaCO₃ in the presence of CO₂ (Neina, 2019) which is normally found at ten times greater levels in soil (0.3 % CO₂) when compared to the CO₂ concentration (0.03%) in the atmospheric air above the soil. Further, PC2 in site year 2 was also identified as qualitative vegetative growth with high loadings of Cu (0.764), which enhances plant photosynthesis; this was also supported by an increased number of branches (0.595) per plant at the time of harvest. PC3 in 2018 and 2019 was identified as improved enzyme activity and pod formation stage, which was supported by high loadings of Fe (0.919 and 0.889), Zn (0.846 and 0.735) and number of pods (0.736 and 0.605) at harvest.

In 2018 (site year 1), PC4 was identified as root proliferation stage, with high loadings of P₂O₅ (0.801) and surface soil moisture (0.727). Root interception of inorganic phosphorous increases with root proliferation. The diffusion of P₂O₅ from the soil through the roots increases with increased moisture content of dry soil (Lambers et al., 2006). Low rainfall (Figure 2) events during critical growth stages of pigeon pea (September 2018 to December 2018), led to dry soil and reduced the dissolution of readily available (water-soluble) nutrients, which are required for plant growth. During the drought, when soil moisture slightly increased from an uneven and low precipitation pattern during November and December 2018, the plant enhanced its root proliferation in the presence of high available soil P₂O₅, in search of soil nutrients (Basu et al., 2016; Peret et al., 2014). In 2019 (site year 2), PC4 was identified as humification with high loadings of EC (0.838), OC (0.558) and N (0.557) suggesting activation of humification (Doni et al., 2014). In site year 1, PC5 was identified as chemical changes in soil, and soil aeration supported by high loadings of soil pH (0.752) and silt (0.674). During site year 2, in PC5, soil ESP (-0.657) and silt (-0.583) showed a high negative factor loading, however available surface moisture (0.754) exhibited a high positive factor loading; this implies that when in water, dissociated calcium cations from CaCO₃ in the presence of CO_2, have replaced sodium ions and adsorbed on the soil colloids (van de Graaff & Patterson, 2001). In site year 1, PC6 had high negative factor loadings of CaCO₃ (-0.781) and positive factor loadings of available boron (0.788), which are attributed to plant cell wall synthesis and root nodule formation for N₂ fixation in leguminous crops. Growing plant cells responds well to boron for dividing and synthesizing cell walls when compared to growth-limited plant cells, in which the synthesis of primary cell walls is negligible (Fleischer et al., 1998). In leguminous plant roots, the development of symbiosis with soil Rhizobium bacteria depends on the concentration of available soil boron and calcium and that both nutrients are essential for nodule structure and fixing atmospheric nitrogen in the plant roots (Redondo-Nieto et al., 2003). As CaCO₃ is a source of calcium required by the plants during critical growth stages, the dissolution of CaCO₃ in soil will be very rapid in the presence of carbon monoxide (CO), (Gallagher & Breecker, 2020), therefore the negative factor loading of CaCO₃ was observed in PC6 of site year 1. In PC6 of site year 2, high factor loadings of available P₂O₅ (-0.715) and sulphur (0.719) were found, whereas in PC7 of site year 1, available copper (0.682) and sulphur (-0.688) had high loadings associated with enhanced enzyme activity essential for nodule formation and nitrogen fixation in the leguminous plant roots (Weisany et al., 2013). PC7 of site year 2 showed a high negative factor loading of available boron (-0.726). Legume crops required more amount of boron compared to most field crops as boron plays a vital role in the development of reproductive organs, pollen germination, flower and fruit setting. Boron being a required trace element, there is a narrow gap between deficiency and toxicity in soil–plant systems (Chatterjee & Bandyopandhyay, 2017). Therefore, one spray at the initiation of the reproductive phase is sufficient for optimum flowering and pod yield (Subasinghe et al., 2003); without initial boron spray, plant uptake will deplete boron in the soil, hence the factor loading in PC7 being negative for boron.

All seven PCs of site year 1, when subjected to stepwise linear regression with crop yield, explained the variance with an R² = 0.68 (p<0.01). The F test of ANOVA was significant at the 1% level. The coefficients of PC3 and PC4 were significant at the 1% level independently, whereas PC7 was significant at the 5% level. Therefore, the variables explaining PC3, PC4 and PC7 such as Fe, Zn, number of pods at harvest, available P₂O₅, soil moisture, and copper significantly influenced the yield (Equation 3, Table 5). Available Zn and P₂O₅ help in seed formation, whereas available soil moisture supports crop growth, which was observed in terms of number of pods per plant and seed yield at the time of harvest. (Subrahmaniyan et al., 2008).

Equation 3:

Y_seed = 1300.913 + 0.636^**PC3 + 0.397^**PC4 + 0.341^*PC7, F=13.414^*, R²=0.679

All seven PCs of site year 2, when subjected to stepwise linear regression with crop yield, explained the variance significantly at the 1 % level ((R² = 0.677 p<0.01). The PC1 and PC3 coefficient was significant at the 1% level. PC6 coefficient was significant at 5 % level (p<0.05). Therefore, the variables constituting PC1, PC3 and PC6 such as K₂O, S, Fe, Zn, Mn, Ca, Mg, CEC, clay, available water content and number of pods at harvest were significantly influencing the yield (Equation 4, Table 5). This may be due to increased availability of nutrients such as K₂O, Fe, Zn, Mn, and the presence of high clay content, which enhances the available water and water holding capacity of the soil; this facilitates nutrient uptake, leading to the increase in number of pods (Rana & Badiyala, 2014; Zarei et al., 2011).

Equation 4:

Y_seed = 1421 + 0.698^**PC₁ + 0.339^**PC3 - 0.275^*PC6, F = 13.264^**, R² = 0.677