Evaluation of spatial distribution and regional zone delineation for micronutrients in a semiarid Deccan Plateau Region of India

2.1 Details of study area

The investigation was carried out in Telangana state of India. The state (lying in 15.83–19.75 N latitude and 77.42–81.75 E longitude) composes of 11.48 million ha area of which nearly 43% area is under cultivation. The state experiences semiarid hot and dry climate. The north zone of the state receives mean annual rainfall of 810–1,135 mm rainfall whereas, south zone receives mean annual rainfall of 560–970 mm (Satyavathi & Reddy, 2004). Maximum rainfall (80% of total precipitation) occurs in the months of June to September. Summer in the state starts in March and peaks in May with mean temperature of 42 °C. Winter starts in late part of November month and extends up to early part of February month with mean temperatures ranging from 22 to 23 °C. Pedologically, the predominant parent materials of soils of the state are igneous and metamorphic rocks (Satyavathi & Reddy, 2004). Entisols, Alfisols, Inceptisols, Vertisols, and Mollisols are the major soil orders of the state and soils are of sandy loam to clayey in texture (Reddy, Shiva Prasad, & Harindranath, 1996). The main crops, grown with tillage, crop residue management and fertilizer application practices, of the state are rice (Oryza sativa), maize (Zea mays), sorghum (Sorghum bicolor), soybean (Glycine max), castor (Ricinus communis), groundnut (Arachis hypogaea), green chillies (Capsicum annum), cotton (Gossypium arboreum), and pulses.

2.2 Sampling of soil and their analysis

Altogether, 4,939 soil samples (with geographical coordinates) were collected from 0 to 0.15 m soil depth with the help of stainless steel auger and following stratified random sampling procedure (Cressie & Chan, 1989) during 2014 and 2016 (Figure 1). Soil samples were collected from agricultural land, predominantly cultivated with field crops, of small (<1 ha), medium (1–3 ha), and large (>3 ha) land holdings. Two to three, five to six, and eight to ten subsamples were collected for making a composite sample from small, medium, and large holdings, respectively. Composite soil samples were prepared to reduce the local noise/sampling effect and to improve the accuracy of prediction (Kerry, Oliver, & Frogbrook, 2010; Webster & Burgess, 1984) as the present study region covers a large area. Processing (air-drying and removal of debris and stones) of soil samples was carried out. Grinding of the samples were carried out to pass through a sieve (2 mm size) before storage in polyethylene container for analysis. Soil properties, pH, and electrical conductivity (EC) were estimated using methods outlined by Jackson (1973) and SOC by Walkley and Black (1934). Available Zn, Fe, Cu, and Mn concentrations in soil samples were extracted by diethylenetriaminepentaaceticacid (C₁₄H₂₃N₃O₁₀) solution (Lindsay & Norvell, 1978) and estimated by atomic absorption spectrophotometer (AAS; VARIAN-Z240, GTA 120).

2.3 Statistical analysis

The parameters of descriptive statistics for soil properties and available micronutrients were obtained using SAS software (SAS Institute, 2011). The normality of dataset was verified by Kolmogorov–Smirnov test (at p < 0.05). Pearson's correlation matrix was obtained for visualizing relationship among soil parameters. Geostatistical analysis of soil properties and available micronutrients was carried out by ArcGIS 10.4.1. Semivariogram for each soil property and available micronutrient (Goovaerts, 1997) was calculated from averaged values. Best fitted model for each soil parameter was selected through the technique of cross-validation. Ordinary kriging was applied for interpolation mapping and kriging biasness and accuracy was tested by cross-validation (Xin-Zhong et al., 2009).

2.4 PC analysis and fuzzy clustering

The values of correlation analysis were used for principal component analysis (PCA) which produced new set of variables called PCs. PCs with >1 eigenvalues were considered for RZ delineation (Davatgar, Neishabouri, & Sepaskhah, 2012). The datasets were partitioned into two to eight clusters by fuzzy c-means clustering technique and FuzME software (Berget, Mevik, & Naes, 2008; Minasny & McBratney, 2006). Normalized classification entropy (NCE) and fuzzy performance index (FPI) were used to obtain optimum cluster number. NCE and FPI accounts for extent of disorganization by specific classes and degree of fuzziness, respectively. According to Fridgen et al. (2004), the highest NCE and the lowest FPI values provided the optimum number of cluster. The variance analysis procedure was adopted to test the differences among the RZs.

(1)

(2)

where c denotes cluster number, n denotes observation number, μik denotes fuzzy membership, and log_a denotes natural logarithm.

3.1 Overall variability of measured soil properties and available micronutrients

Soils were acidic (pH 4.41) to alkaline (pH 10.71) in reaction and nonsaline (EC 0.15 to 1.31 dS/m) in character with the mean value of 7.48 for soil pH and 0.42 dS/m for EC (Table 1). The SOC content varied from 0.16% to 1.09% with average value 0.42%. Our results support the findings of Reddy et al. (1996) and Satyavathi and Reddy (2004) who reported wide ranges for soil pH, EC, and SOC in the region. This may be ascribed to varied soils, prevailing climatic conditions and various crop husbandry practices followed in the region. Available Zn, Fe, Cu, and Mn concentrations varied widely with mean values of 0.83, 8.79, 0.99, and 8.79 mg/kg, respectively. It was found that 26%, 10%, 9%, and 9% areas were deficient (comprising acute deficient and deficient) in Zn, Fe, Cu, and Mn, respectively, as per the critical limits of the study region (Shukla et al., 2016). Our finding is similar to the results reported by Shukla et al. (2016) who recorded the mean values of 1.66, 1.37, 12.20, and 10.30 mg/kg for available Zn, Cu, Fe, and Mn concentrations, respectively in soils of Indian TGP region. Shukla et al. (2017) also reported the mean concentration of 2.24, 1.49, 19.01, and 36.76 mg/kg soil for plant available Zn, Cu, Mn, and Fe, respectively, in the soils of SHR of India. Soil properties exhibited low (only soil pH) to moderate (rest of soil properties) variability with <10, 10 to 100 and > 100% of CV values indicating variability to the extent of low, moderate, and high degree, respectively (Nielsen & Bouma, 1985). Bogunovic et al. (2017) reported low, medium, and high variability for pH, organic matter, and EC, respectively in soils of Rasa river valley of Croatia. Behera and Shukla (2015) reported low (for pH and EC) to moderate (for SOC content) variability in Indian acid soils. Similarly, variability was low for pH and medium for EC and organic matter in soils of Alequeva reservoir of Portugal (Ferreira, Panagopoulos, Andrade, Guerrero, & Loures, 2015). In soils of northern Ethiopia, low (for pH) and medium variability (for SOC and available Fe) were reported (Tesfahunegn et al., 2011). Moderate variability for available Zn, Fe, Cu, and Mn was recorded by Wang, Wu, Liu, Huang, and Fang (2009) in China's paddy growing soils and by Shukla et al. (2016), Shukla et al. (2017) in Indian TGP and SHR soils. However, Foroughifar, Jafarzadeh, Torabi, Pakpour, and Miransari (2013) recorded high variability for available Fe and moderate variability for available Cu, Mn, and Zn in Dasht-e-Tabriz soils of Iran. Among the soil properties, available Zn, Fe, Cu, and Mn had higher CV values than soil pH, EC, and SOC content. High variability in soil micronutrients is ascribed to different micronutrients content of parent material, pedogenic processes, and diversity in weathering regimes (Bowen, 1979).

3.2 Relationship among the soil properties and available micronutrients

Correlation coefficient values in Table 2 reveals the relationship among the soil properties and available micronutrients. Correlation coefficient values indicated negative correlation of pH with soil Zn, Mn, and Fe. The correlation of soil EC with SOC and available Zn, Fe, Cu, and Mn was negative. The correlation of SOC content with available Zn, Cu, and Mn was positive, whereas it was negative with available Fe. Similarly, Wei, Hao, Shao, and Gal (2006) reported negative correlation between pH and plant available Zn, Mn, and Fe in China's Loess Plateau region soils. The same authors recorded positive correlation of soil organic matter with plant available Zn, Mn, and Fe. Shukla et al. (2016) and Shukla et al. (2017) reported negative correlation of pH with available Fe, Zn, Mn, and Cu in TGP and SHR soils of India. Positive correlation of SOC was recorded with all the four cationic micronutrients in soils of TGP and with only Zn and Fe in SHR soils of India. The negative correlation among soil pH and available Fe, Zn, and Mn in our study is obvious as changes in soil pH influence soil micronutrient content (Neilsen, Hoyt, & Mackenzie, 1986). According to Lindsay (1979), with every unit of soil pH enhancement in the range of 4–9, Zn, Cu, and Mn solubility reduces by 100-fold as compared to 1,000-fold reduction for Fe in soil. Positive correlation of SOC with available micronutrient in soil is because of the fact that SOC is a key component of soil organic matter. Soil organic matter enhances nutrient availability to the crop plants by releasing organic substances which can chelate with micronutrients and thereby improving their availability (Tisdale, Nelson, & Beaton, 1985). Positive correlations among the available micronutrients, Zn versus Mn, Zn versus Fe, Fe versus Cu, Fe versus Mn, and Cu versus Mn were recorded. This indicates that similar sets of factors influence distribution of these metallic nutrients in soils in the study region. Behera and Shukla (2013) have reported positive correlation between available Zn with available Cu and Fe and between available Cu with available Fe in some Indian acid soils under cultivation. The observed variation of studied soil properties and available micronutrient and their relationship in DPR of India is attributed to soil types, climatic condition, and practices of crop managment. This warrants proper characterization of soil though accurate sampling, use of geostatistics and appropriate zoning.

3.3 Spatial characters of soil properties and available micronutrients

Soil properties and available micronutrients had different semivariogram characters (Table 3, Figure 2). The best fitted model obtained from semivariogram analysis was stable for soil pH and available Fe, exponential with respect to EC, available Zn, and Cu and K-Bessel with respect to SOC and available Mn. Foroughifar et al. (2013) reported best fitted spherical model for pH, SOC, available Zn, Cu, and Fe and linear best fitted model for available Mn. Likewise, from a study in rice cultivated soils of eastern India, Tripathi et al. (2015) reported best fitted spherical model for soil pH, EC, SOC, and available Zn and Cu, pentaspherical model for available Fe and exponential model for available Mn. The nugget values, which is the measure of variance due to errors, were low (varied from 0.00 to 0.59) for the soil properties beside available Fe (58.51) and Mn (19.17). This is in line with our findings of higher CV values for available Fe and Mn in comparison with other soil properties.

The nature of spatial dependency of soil parameters is described as strong (nugget to sill ratio of ≤0.25), moderate (0.25–0.75), and weak (>0.75) (Cambardella et al., 1994). In the current investigation, SOC (0.000) and available Zn (0.160) recorded strong spatial dependency whereas soil pH (0.543), EC (0.600), available Fe (0.560), Cu (0.700), and Mn (0.469) recorded moderate spatial dependency. Strong spatial dependency for SOC and available Zn is ascribed to soil types and prevailing climatic conditions in the study region. Whereas moderate spatial dependency for other soil properties is attributed to inherent soil characteristics as well as adoption of different cropping sequences and methods of nutrient manipulation. Range value of the semivariogram is the key for deciding the design of sampling for soil properties mapping (Zhang, Zhuang, Qian, Wang, & Ji, 2015), because it is the distance within which the two samples are associated. Higher range value indicates the influence of anthropogenic and natural factors on the soil property for a greater distance (Lopez-Granados et al., 2002). In our study, soil properties and available micronutrients had range values spanned from 9,574 m (for SOC) to 96,024 m (for available Fe) which is similar to the findings of Shukla et al. (2017). Different ranges for soil properties and available micronutrients are ascribed to joint actions of nature of soils, prevailing climate and anthropogenic activities for managing land and crops.

Further, ordinary kriging technique was adopted to generate distribution variability maps for soil properties and available micronutrients (Figure 3). Heterogeneous distribution pattern for soil properties and available micronutrients across the region was observed. Soil pH map indicated that about 47% of study area falls under the category of near neutral (6.5–7.5), which is conducive for cultivation of most of the field crops. About 7% and 5% of the study area was having soils with pH < 6.5 and pH > 8.5, respectively (Figure 3). These areas require immediate attention of the land managers. The locality had very low pH (pH < 5.5) typical of the high altitude hilly terrain. Low soil pH of the study area is ascribed to the parent materials such as granite, gneiss, schist, and ferruginous from which these soils are developed. These acidic parent materials weathered to form red sandy soils and red loam soils. Though the EC values varied in the study area and the map exhibited varied distribution pattern, soil was not saline as the values of EC were < 2 dS/m. About 53% of the area had low (<0.50%) SOC content and found distributed in southern and west-northern part of the study area. Forty-three percent of the area had medium (0.50–0.75%) SOC content, and mostly found in eastern and southern part. Low content and varied distribution of SOC in the area is because of prevailing arid climate, physiography, land use pattern, localized microclimate, and several biological processes such as biomass production, deposition, and decomposition of litter (Mao et al., 2015). Distribution maps exhibited different pattern of distribution for available micronutrients. Higher quantity of Cu and Mn was recorded in south and south-eastern part of the study area. About 26%, 19%, 8%, and 4% of the study area were having <0.6, <5.5, <0.4, and < 2.0 mg/kg of Zn, Fe, Cu, and Mn concentrations, respectively. This is in line with findings of Shukla et al. (2015). Different concentrations of available micronutrients in the area is attributed to nature of soils and pedogenic and geochemical processes. The study area possesses igneous and metamorphic rocks with varied micronutrient concentration as parent material, which influence the micronutrient availability.

3.4 Delineation of RZs

Because most of the soil properties were correlated significantly, PCA was carried out. First four PCs (with >1 eigenvalue and explaining 73% variability) were used for the study (Kaiser, 1960) (Table 4, Figure 4). Twenty-six percent of the total variance was explained by first PC (PC1), which was dominated by SOC, available Fe and Mn. Second PC (PC2) was mostly affected by soil pH and EC, and it explained 19% of variability. The third and fourth PCs (PC3 and PC4) explained an additional 15% and 13% of the variance and was predominated by available Zn and Cu, respectively. PCA aggregated seven properties into four PCs explaining most of spatial variability. Similarly, by aggregating the variability of the studied soil properties, Khaledian, Kiani, Ebrahimi, Brevik, and Aitkenhead-Peterson (2017) had recorded four PCs in their study area. Biplot analysis of PC1 versus PC2 revealed two prominent groups of soil properties and available micronutrients, in which SOC and available Cu and Mn constituted one group and EC and available Zn and Fe created another (Figure 5).

The first four PCs were considered for developing RZs. Clustering of the four PCs resulted into six clusters for the area by considering NCE and FPI functions (Figure 6). The developed maps exhibiting six RZs were different from each other and are presented in Figure 7. Yan, Zhou, Feng, and Hang-Yi (2007), Davatgar et al. (2012) and Shukla et al. (2017) also reported different soil MZs in their respective study area. The RZ 6 was having highest geographical area (22.5%) followed by RZ 3 (21.6%), RZ 4 (19.8%), RZ 2 (14.6%), RZ 5 (13.0%), and RZ 1 (8.5%). Soil properties of the RZs varied significantly (Table 5). Lowest soil pH was recorded in RZ 4 and RZ 5 and highest in RZ 6 while RZ 3 was having lowest SOC content. Lowest available Zn concentration was recorded in RZ 3 and RZ 6. Though the average concentration of available Fe, Cu, and Mn in different RZs were higher than the critical limits adopted for the region, there was different levels of deficiency in various parts of the RZs. The heterogeneity in soil properties including available Zn, Fe, Cu, and Fe in six RZs is owing to nature of soils, agro-ecological factors, climate, and different cultural practices adopted for managing nutrients and crops. It gives an insight for effective soil management specific to different sites for sustainable soil and crop productivity for ensuring food security and reducing land degradation. The mean values of soil properties and available micronutrients in different RZs should be used as guide for effective management of soil micronutrients needed for specific sites.

For example, RZ 2 to RZ 6 had low Zn availability, in which increased Zn supply will help in improving the crop productivity. Shukla et al. (2015) reported increased rice productivity in Nalgonda (in MZ 2), Nizamabad and Rangareddy (in RZ 3), and Karimnagar (in RZ 5) Districts of the study area where in Zn was applied. Productivity of other crops such as sorghum, groundnut, and green chillies improved after Zn fertilization in these areas. Likewise, immediate attention is required to enhance the content of SOC content by various management practices such as conservation tillage, organic manuring, and crop rotation with leguminous crops in RZ 3 having the lowest SOC. Hence, there is need to devise different soil management strategies for various RZs based on soil properties and available micronutrient values. The land manages having sufficient information about the soil properties and available micronutrients could better utilize the RZs if they devise simple, easy to adopt and economically feasible management strategies for different regions for enhanced crop production.