Integration of geostatistics, multivariate analysis and geographically weighted fuzzy clustering for delineating soil fertility management zones in rice-growing regions of Telangana

Abstract

Introduction:

Spatial variability of soil fertility limits efficient nutrient management and sustainable rice production. Delineating management zones (MZs) using spatial soil information provides a practical framework for site-specific nutrient management (SSNM). This study developed an integrated approach combining geostatistics, principal component analysis (PCA), and geographically weighted fuzzy clustering (GWFC) to delineate soil fertility management zones in rice-growing regions of Telangana, India.

Methods:

A total of 200 geo-referenced surface soil samples (0–15 cm) were analyzed for soil pH, electrical conductivity (EC), organic carbon (OC), available nitrogen (N), phosphorus (P₂O₅), and potassium (K₂O). Geostatistical analysis and kriging interpolation characterized spatial variability, while PCA identified the major factors influencing soil fertility. GWFC integrated soil attributes and spatial proximity to delineate homogeneous management zones. The optimal number of zones was determined using the Partition Coefficient (PC), Fuzzy Performance Index (FPI), Normalized Classification Entropy (NCE), and Xie–Beni (XB) indices.

Results:

All soil properties exhibited strong spatial dependence, with nugget-to-sill ratios below 13.28% and spatial ranges of 662–1592 m. PCA extracted two principal components explaining 95.94% of the total variance. Three optimal management zones differed significantly (p < 0.05) in soil fertility and crop productivity. Site-specific fertilizer recommendations based on Soil Test Crop Response (STCR) equations reduced fertilizer application by up to 38.6% for N, 100% for P₂O₅, and 50.0% for K₂O. Grain yield increased from 53 to 67 q ha⁻¹, net profit improved by up to 85%, and the benefit–cost ratio increased from 1.66 to 2.40.

Discussion:

The integrated GWFC framework effectively delineated management zones and captured spatial soil fertility variability. The proposed approach provides a reliable decision-support tool for precision nutrient management, improving fertilizer-use efficiency, productivity, profitability, and the sustainability of rice-based production systems.

1 Introduction

Semi-arid regions are characterized by high temperatures, erratic rainfall patterns, and frequent moisture stress, all of which impose significant constraints on sustainable agricultural production. Soils in these environments exhibit considerable spatial and temporal variability due to the combined influences of parent material, topography, climate, land-use history, and management practices (1). Furthermore, prolonged land degradation, inappropriate nutrient management, and climatic stress have adversely affected soil health and crop productivity across many semi-arid agroecosystems (2). Low levels of soil organic matter reduce aggregate stability, impair nutrient retention, and increase susceptibility to erosion and surface crusting, thereby limiting crop growth and nutrient availability (2). As a result, nutrient use efficiency (NUE) in semi-arid agricultural systems often remains below 25%, leading to poor fertilizer-use efficiency and inconsistent crop responses to nutrient application (3).

Rice-based cropping systems in the semi-arid regions of India are particularly vulnerable to these challenges. High evaporative demand, irregular rainfall distribution, and substantial heterogeneity in soil fertility create considerable variability in crop growth and nutrient requirements within individual fields. Such variability often results in localized zones of nutrient deficiency or nutrient accumulation, reducing overall fertilizer-use efficiency and crop productivity (4). Despite this inherent variability, fertilizer recommendations are generally applied uniformly across fields, assuming homogeneous soil conditions. This blanket approach frequently leads to nutrient over-application in some areas and under-application in others, resulting in reduced productivity, increased production costs, and greater environmental risks associated with nutrient losses through leaching and runoff (5). Therefore, improving nutrient management strategies that account for spatial variability has become essential for enhancing the productivity, profitability, and sustainability of rice-based production systems. Precision agriculture offers a scientifically robust framework for managing spatial variability by matching agricultural inputs to site-specific soil and crop requirements. Among the various precision agriculture strategies, the delineation of management zones (MZs) has emerged as an effective approach for addressing within-field heterogeneity (6). Management zones are defined as sub-field areas exhibiting relatively homogeneous soil properties, productivity potential, and nutrient requirements, thereby enabling the implementation of site-specific nutrient management (SSNM) practices (7, 8). Numerous studies have demonstrated that management-zone-based nutrient management can improve fertilizer-use efficiency up to 30%, reduce input costs by 10–30%, and maintain or enhance crop yields (2, 9, 10). However, scientific evidence regarding the delineation and validation of management zones for rice-based cropping systems in the semi-arid regions of India remains limited.

Accurate management zone delineation requires analytical techniques capable of capturing the complex spatial variability and interrelationships among soil properties. Traditional approaches based on soil surveys, topographic information, yield maps, or individual nutrient indicators often fail to adequately represent the multivariate and spatially dependent nature of soil fertility variability (11–13). Soil properties are inherently spatially correlated, and their distribution is influenced by interactions among environmental conditions, hydrological processes, and management interventions. Consequently, identifying agronomically meaningful management zones requires methods that simultaneously account for both soil attribute variability and spatial relationships among sampling locations (14–16).

Recent advances in geostatistics and geospatial technologies have significantly enhanced the ability to quantify and map soil variability at fine spatial resolutions (10, 17). Geostatistical methods, particularly semivariogram analysis and kriging interpolation, provide robust tools for characterizing spatial dependence and generating continuous spatial distributions of soil properties (18). Likewise, multivariate statistical techniques such as principal component analysis (PCA) facilitate the identification of key factors controlling soil fertility variability and reduce data dimensionality while retaining essential information (19). These approaches provide valuable insights into the underlying processes governing soil heterogeneity and serve as important inputs for management zone delineation. Clustering techniques have become increasingly popular for delineating management zones because they can group locations with similar soil and environmental characteristics. Among them, fuzzy C-means (FCM) clustering has been widely applied in precision agriculture due to its ability to accommodate gradual transitions between classes and represent the continuous nature of soil variability (9, 20). However, conventional FCM clustering primarily considers attribute similarity and does not explicitly incorporate spatial location into the clustering process. As a result, the delineated zones may not adequately capture local spatial patterns and may produce fragmented clusters that are difficult to implement in field-scale management.

Geographically Weighted Clustering (GWC) overcomes this limitation by integrating spatial proximity with soil attribute similarity during cluster formation. By incorporating local spatial relationships, GWC generates clusters that better reflect the underlying spatial processes governing soil variability and agricultural productivity (21). This characteristic is particularly important in rice-growing environments where soil fertility patterns are strongly influenced by micro-topography, water movement, and localized management practices. Consequently, geographically weighted clustering has the potential to generate more realistic and operationally useful management zones than conventional clustering approaches. Despite the growing application of clustering techniques in precision agriculture, several research gaps remain. First, most studies in rice-based production systems have relied on conventional clustering approaches that do not adequately account for local spatial heterogeneity. Second, relatively few studies have integrated geographically weighted clustering with objective cluster validation techniques to determine the optimal number and quality of management zones. Third, limited information is available regarding the application of such integrated approaches for soil fertility zoning and site-specific nutrient management in the rice-growing regions of semi-arid India. Consequently, the practical utility of spatially weighted clustering approaches for improving fertilizer recommendations and nutrient-use efficiency remains insufficiently explored.

The novelty of the present study lies in the integration of geostatistical analysis, principal component analysis, geographically weighted clustering, and multiple fuzzy clustering validity indices, including the Partition Coefficient (PC), Fuzzy Performance Index (FPI), Normalized Classification Entropy (NCE), and Xie–Beni (XB) Index, within a unified framework for management zone delineation. Unlike conventional approaches, the proposed methodology simultaneously accounts for spatial dependence, multivariate soil variability, and cluster validation, thereby improving the accuracy, reliability, and practical relevance of management zone delineation for precision nutrient management.

Therefore, the objectives of this study were to: (i) characterize the spatial variability and spatial dependence of key soil physicochemical properties using geostatistical techniques; (ii) identify the dominant factors controlling soil fertility variability through principal component analysis; (iii) delineate optimal soil fertility management zones using geographically weighted clustering supported by PC, FPI, NCE, and Xie–Beni validity indices; and (iv) evaluate the implications of management-zone-based nutrient management for improving fertilizer-use efficiency, profitability, and sustainability in rice-based cropping systems of Telangana, India.

2 Materials and methods

2.1 Study area description

The present study was conducted in Ganturpally village, Hasanparthy Mandal, Hanamkonda district, Telangana, India (18°5’58.29”N to 18°6’13.74”N, 79°30’0.92”E to 79°31’53.68”E) (Figure 1). The study area lies within the Deccan Plateau agro-climatic region and is characterized by a hot semi-arid climate. The duration of the rice-growing season generally ranges from 100 to 120 days (22). Annual rainfall is unevenly distributed, with the majority received during the southwest monsoon and the remaining portion during the northeast monsoon. Mean summer temperatures range from 31 to 39 °C, while winter temperatures typically vary between 14 and 25 °C (23).The soils of the study area are deep, black, calcareous, and fine-textured, exhibiting a neutral to strongly alkaline reaction. They are characterized by a medium to strong subangular blocky structure. According to the USDA Soil Taxonomy, the soils are classified as very fine, smectitic, hyperthermic Typic Chromusterts (24, 25). The dominant cropping system is rice in kharif (the monsoon season, June-October) followed by rice in rabi (the post monsoon/winter season, November-March). Recommended fertilizer doses are 120:60:40 kg N:P₂O₅:K₂O ha^-1 for kharif rice and 150:60:40 kg N:P₂O₅:K₂O ha^-1 for rabi rice. Irrigation is primarily supplied through groundwater.

Figure 1

2.2 Collection and analysis of soil samples

Soil sampling was conducted during 2024 in the agricultural lands of Ganturpally village, Hasanparthy Mandal, Hanamkonda district, Telangana. A total of 200 geo-referenced soil samples were collected from the 1000 ha study area within 0–15 cm depth following a stratified sampling approach, where sampling locations were selected to represent variability in land use, topography, and soil type rather than a fixed grid or systematic interval. Previous studies have indicated that approximately 100–150 samples are sufficient for reliable variogram estimation (26). Therefore, the use of 200 samples in the present study ensures robust spatial representation and improves the reliability of variogram modeling, kriging, and subsequent clustering analysis. The stratified sampling approach offers several advantages, including improved representation of spatial variability, reduced sampling error, and enhanced efficiency in heterogeneous agricultural landscapes. This design is particularly suitable for precision agriculture studies where soil properties are influenced by multiple interacting factors. Approximately 500 g of soil was collected from each sampling location. The geographic coordinates of all sampling points were recorded using a Garmin handheld Global Positioning System (GPS) receiver device with a positional accuracy of approximately ±3–5 m to enable spatial verification and subsequent geospatial analysis. Following collection, soil samples were air-dried under shade, gently crushed using a wooden hammer to minimize contamination, and thoroughly homogenized. The processed samples were sieved through a 2 mm mesh for routine chemical analysis. A separate subsample was further passed through a 0.2 mm sieve for the determination of soil organic carbon. Soil pH and electrical conductivity (EC) were measured in a 1:2.5 soil-to-water extract using a digital pH meter (Orion Star A211, Thermo Fisher Scientific, USA) and a conductivity meter (Orion Star A212, Thermo Fisher Scientific, USA), respectively, following standard analytical procedures (25). Soil organic carbon (SOC) was estimated by the wet oxidation method using potassium dichromate and expressed as percentage (%) (27). Available nitrogen (N) was determined using the alkaline potassium permanganate (KMnO₄) method (28) with a Kel Plus Nitrogen Analyzer (Pelican Equipments, India). Available phosphorus (P₂O₅) was extracted using 0.5 M sodium bicarbonate solution (pH 8.5) following the Olsen method (29) and quantified using a double-beam UV–Visible spectrophotometer (Shimadzu UV-1800, Japan) at a wavelength of 660 nm. Available potassium (K₂O) was extracted with 1 N ammonium acetate (30) and measured using a flame photometer (Elico CL 378, India). Available nitrogen, phosphorus, and potassium contents determined from the surface soil layer (0–15 cm) were expressed as kg ha^-¹. Phosphorus and potassium are reported as P₂O₅ and K₂O equivalents, respectively, following conventional Indian soil fertility evaluation practices and the requirements of the STCR-based fertilizer recommendation equations used in this study.

2.3 Statistical analysis

Descriptive statistical analyses were performed to characterize the distribution and variability of soil physicochemical properties across the study area. The computed statistical parameters included minimum, maximum, mean, median, standard deviation (SD), coefficient of variation (CV), skewness, and kurtosis. These indices provided information on the central tendency, dispersion, and distributional behavior of the dataset. The coefficient of variation was used to quantify the degree of spatial variability among soil properties, while skewness and kurtosis were employed to assess deviations from normality. The results of the descriptive analysis served as a preliminary assessment of data quality and distribution prior to conducting correlation, geostatistical, and multivariate analyses. All statistical computations were carried out using RStudio (version 2025.05.1 + 513).

2.4 Geostatistical analysis

The overall workflow adopted for geostatistical analysis and soil variability assessment is summarized in Figure 2. Geostatistics deals with the analysis of variables that exhibit spatial or temporal dependence and provides tools to characterize spatial continuity while incorporating this dependence into prediction and mapping procedures. By exploiting spatial autocorrelation, geostatistical approaches enable improved estimation compared to conventional statistical methods (15, 16).

Figure 2

Spatial dependence of soil properties was examined using semivariogram analysis, which represents the degree of similarity between observations as a function of separation distance. The semivariogram provides a quantitative measure of spatial continuity and is constructed by computing the average of half the squared differences between paired observations separated by a given lag distance and direction. Experimental semivariograms were generated for each soil variable to evaluate spatial structure and correlation patterns (19, 26). The experimental semivariance, was estimated as Equation 1 (9, 10, 31):

Where:

= experimental semivariance at lag distance ;

= number of pairs of observations separated by distance ;

= measured value of the regionalized variable at location ;

= measured value at a location separated from by lag distance .

Experimental semivariograms were computed from the observed semivariance pairs for different lag distances. Theoretical semivariogram models (spherical, exponential, and Gaussian) were fitted to the experimental semivariograms using weighted least-squares estimation to derive nugget, sill, and range parameters. Model accuracy was assessed through leave-one-out cross-validation (LOOCV) implemented in ArcGIS Geostatistical Analyst, and the model with the lowest root mean square error (RMSE) and acceptable prediction performance was selected for ordinary kriging and spatial interpolation.

Semivariograms were developed using ArcGIS10.8 (ESRI) software, and theoretical models were fitted to the experimental semivariograms to describe spatial autocorrelation. The spherical, Gaussian, and exponential models were evaluated to determine the most appropriate representation of spatial structure for each soil attribute (21, 32).The spatial structure of each soil property was evaluated using the nugget-to-sill ratio C₀/(C₀ + C₁), which reflects the proportion of spatially unstructured variance. Based on this ratio, the degree of spatial dependence was classified as strong (<25%), moderate (25–75%), or weak (>75%) (33). The experimental semivariograms were fitted with theoretical models, and the model providing the best predictive performance was selected using cross-validation statistics. The accuracy of prediction was evaluated using the root mean square error (RMSE) is calculated using Equation 2 (34, 35).

Where:

= root mean square error;

= total number of observations used in cross-validation;

= measured (observed) value at the location;

= predicted value at the location.

Lower RMSE values indicate better predictive performance of the semivariogram model. Therefore, the model with the lowest RMSE was selected for interpolation of each soil property. The validated semivariogram models were subsequently used for spatial interpolation through ordinary kriging. Ordinary kriging estimates values at unsampled locations by assigning optimal weights to neighboring observations based on the spatial structure defined by the semivariogram. The ordinary kriging estimator is expressed by Equation 3.

Where:

= estimated value at the unsampled location ;

= observed value at the sampled location ;

= kriging weight assigned to the observation;

= number of neighboring observations used in the estimation.

The kriging weights were determined from the fitted semivariogram model to ensure unbiased estimation with minimum estimation variance. Spatial distribution maps of soil properties were generated in ArcGIS using kriging interpolation and visualized using appropriate color gradients and contour representations.

2.5 Principal component analysis

Principal component analysis (PCA) (Figure 3) was used to reduce data dimensionality and minimize multicollinearity among soil variables by transforming correlated parameters into a smaller set of uncorrelated principal components. Prior to principal component analysis (PCA), all soil variables were standardized using z-score transformation to remove the effects of differing units and scales. PCA was subsequently performed on the correlation matrix of the standardized variables, ensuring that each variable contributed equally to the analysis regardless of its original measurement scale. This approach is widely recommended when variables are expressed in different units and exhibit varying magnitudes of variance (36). Principal components with eigenvalues equal to or greater than one were retained for further analysis and subsequently used for management zone delineation. Factor loadings were examined to identify the soil properties contributing most to each component, facilitating interpretation of spatial variability (18, 37–39). PCA were carried out in RStudio (version 2025.05.1 + 513) software.

Figure 3

2.6 Geographically weighted clustering analysis

Geographically Weighted Clustering Analysis (GWCA) was employed to delineate management zones by simultaneously considering soil attribute similarity and spatial proximity among sampling locations. Unlike conventional clustering methods, which group observations solely based on attribute similarity, GWCA incorporates geographic coordinates into the clustering process, thereby accounting for local spatial heterogeneity (40). This enables the identification of spatially coherent clusters that better reflect underlying soil-forming factors, management practices, and environmental gradients (41). The analysis was performed using standardized soil variables, including pH, electrical conductivity (EC), organic carbon (OC), available nitrogen (N), phosphorus (P₂O₅), and potassium (K₂O), together with their corresponding geographic coordinates. Prior to clustering, all variables were standardized using z-score transformation to eliminate scale effects and ensure equal contribution of each variable to the clustering process. A geographically weighted fuzzy c-means (GW-FCM) approach was adopted, in which spatial weights were incorporated into the clustering algorithm through a Gaussian kernel function. The spatial weighting scheme assigns greater influence to nearby observations than to distant ones, allowing local spatial relationships to influence cluster formation. An adaptive bandwidth based on the 15 nearest neighbors was used to accommodate variations in sampling density across the study area.

The optimal number of management zones was determined by evaluating clustering performance for different cluster solutions using four widely accepted fuzzy clustering validity indices: Partition Coefficient (PC), Fuzzy Performance Index (FPI), Normalized Classification Entropy (NCE), and Xie–Beni (XB) Index. The most suitable clustering solution was selected based on maximum partition stability and minimum classification uncertainty, resulting in management zones that exhibited high within-zone homogeneity and strong between-zone separation.

The GW-FCM algorithm was implemented in RStudio (version 2025.05.1 + 513) using the packages sp (version 2.2-0) for spatial data handling, geosphere (version 1.5-20) for distance calculations, GWmodel (version 2.4-1) for geographically weighted analysis, and cluster (version 2.1.8.1) for clustering procedures. Custom R scripts were developed to integrate spatial weighting with fuzzy clustering and to calculate clustering validity indices. The algorithm was executed with a maximum of 100 iterations and a convergence threshold of 1 × 10^-5.

2.7 Determination of the optimal number of management zones

The optimal number of management zones was determined using established fuzzy cluster validity indices, namely the Partition Coefficient (PC), Fuzzy Performance Index (FPI), and Normalized Classification Entropy (NCE) (15, 42–44). These indices evaluate the quality of clustering by quantifying cluster compactness, separation, and the degree of fuzziness in cluster membership, thereby providing an objective basis for selecting the appropriate number of zones.

2.7.1 Partition Coefficient

The Partition Coefficient (PC) measures the degree of membership sharing among clusters and reflects the overall sharpness of cluster partitioning and calculate using Equation 4:

where is the membership degree of observation i in cluster k, n is the number of observations, and K is the number of clusters. PC values range from to 1, with higher values indicating crisper and more distinct clustering. However, PC tends to increase monotonically with a decreasing number of clusters and is therefore interpreted in conjunction with other indices.

2.7.2 Fuzzy Performance Index

The Fuzzy Performance Index (FPI) evaluates the extent of fuzziness in the classification and is sensitive to the overlap between clusters and calculated using Equation 5:

FPI values range between 0 and 1, where lower values indicate reduced fuzziness and better cluster separation. The optimal number of management zones corresponds to the minimum FPI value, reflecting the most meaningful partitioning of the dataset.

2.7.3 Normalized Classification Entropy

The Normalized Classification Entropy (NCE) quantifies the uncertainty associated with fuzzy membership assignments and measures the randomness in cluster allocation. It is calculated using Equation 6:

NCE values range from 0 to 1, with lower values indicating less ambiguity and more definitive cluster membership. The optimal number of clusters is identified at the point where NCE reaches its minimum, representing improved cluster clarity.

2.7.4 Xie–Beni

The Xie–Beni (XB) Index evaluates clustering quality by simultaneously considering cluster compactness and cluster separation. It is particularly effective in fuzzy clustering because it incorporates both fuzzy membership values and distances between observations and cluster centers and calculated using Equation 7:

Where,

is the membership degree of observation in cluster ,

is the fuzziness exponent,

is the observation vector,

and are the centers of clusters and ,

is the number of observations, and is the number of clusters.

The numerator represents within-cluster dispersion, weighted by fuzzy memberships, while the denominator represents the minimum squared distance between cluster centers, indicating inter-cluster separation. Lower XB values indicate more compact clusters with better separation, reflecting improved clustering performance. The optimal number of management zones is identified at the minimum XB value, which corresponds to the most balanced partitioning of the dataset in terms of homogeneity within zones and heterogeneity between zones. Due to its ability to penalize overlapping clusters and poorly separated centroids, the XB index is considered one of the most robust validity measures for delineating management zones in precision agriculture studies.

2.8 Selection of optimal management zones

The optimal number of management zones was determined by jointly evaluating PC, FPI, and NCE across a range of cluster numbers. The number of clusters corresponding to maximum PC and minimum FPI and NCE was selected as optimal, ensuring a balance between cluster compactness, separation, and interpretability (Figure 4). This multi-index approach enhances the robustness of management zone delineation and supports reliable site-specific soil and nutrient management decisions. To evaluate differences in soil properties among the MZs, a one-way ANOVA was performed using SPSS (version 26.0) software (19, 44–47). The final MZ maps were generated in ArcGIS 10.8 (48). A schematic representation of the methodological workflow employed in this study is shown in Figure 5.

Figure 4

Figure 5

2.9 Development of site-specific fertilizer recommendations

Site-specific fertilizer recommendations for nitrogen (N), phosphorus (P₂O₅), and potassium (K₂O) were developed for each management zone using the Soil Test Crop Response (STCR) targeted yield equations developed under the All India Coordinated Research Project (AICRP) on STCR, Professor Jayashankar Telangana State Agricultural University (PJTSAU), Telangana, India. The fertilizer prescription equations were used to estimate the nutrient requirement based on the targeted yield approach by considering soil available nutrient status and crop nutrient demand. The fertilizer requirement was calculated using the following Equations 8–10:

Where, FN, FP₂O₅ and FK₂O represent the fertilizer doses of nitrogen, phosphorus and potassium (kg ha^-¹), respectively; T denotes the targeted crop yield (t ha^-¹); and SN, SP and SK indicate soil available nitrogen, phosphorus and potassium (kg ha^-¹), respectively. The equations were applied separately for each management zone using the respective soil fertility values derived from spatial analysis. Based on the calculated fertilizer doses, management zone-specific nutrient recommendations were developed to optimize fertilizer use efficiency and reduce excessive nutrient application.

2.10 Economic analysis of management zone-based nutrient management

The economic performance of management zone-based fertilizer recommendations was evaluated by comparing grain yield, cost of cultivation, gross returns, net returns, and benefit-cost (B) ratio under different management zones with the conventional farmer fertilizer practice. Grain yield (q ha^-¹) was recorded at crop harvest and standardized to hectare basis. The cost of cultivation was estimated by considering all input costs, including fertilizers and other recommended crop management operations. Gross returns (₹ ha^-¹), net returns (₹ ha^-¹), and benefit-cost (B:C) ratio were calculated using Equations 11–13, respectively.₹

Economic analysis was performed following standard farm management procedures described by Gomez and Gomez (49) and CIMMYT (50).

2.11 Sustainability indicators analysis

Sustainability indicators were selected to represent agronomic performance, economic returns, and resource-use efficiency under management zone-based nutrient management. The indicators included grain yield, net profit, benefit-cost ratio, nitrogen saving, phosphorus saving, and potassium saving. These variables were selected because they collectively reflect productivity, profitability, and fertilizer-use efficiency, which are considered key dimensions of sustainable nutrient management (51, 52). To facilitate comparison among indicators measured in different units and magnitudes, all variables were normalized using the min–max normalization technique. The normalized indicator values were computed using Equation 14 before generating the graphical representations.

where:

= observed value, = minimum value, = maximum value.

This procedure transformed all indicators to a dimensionless scale ranging from 0 to 1 and allowed direct comparison among variables with different units. The normalized values were used to construct radar plots for visual comparison of the overall sustainability performance of management zones and conventional farmer fertilizer practice. Similar normalization approaches for multi-indicator assessment have been widely used in agricultural sustainability studies (53, 54).

2.12 Sustainability Performance Index

A Sustainability Performance Index (SPI) was developed to provide an integrated assessment of agronomic performance, economic returns, and nutrient-use efficiency under different management zones. Six indicators, namely grain yield, net profit, benefit-cost ratio, nitrogen saving, phosphorus saving, and potassium saving, were selected because they represent the major dimensions of sustainable crop production, including productivity, profitability, and efficient utilization of fertilizer resources (52, 54). Prior to the development of the sustainability index, all indicators were standardized using the min–max normalization technique to eliminate differences in measurement units and transform the variables onto a common scale ranging from 0 to 1. The normalized value of each indicator was calculated using Equation 14.

The Sustainability Performance Index (SPI) for each treatment was calculated as the arithmetic mean of the normalized indicator values using Equation 15:

where:

SPI = Sustainability Performance Index,

= normalized value of the ith indicator,

n = total number of indicators included in the index.

Higher SPI values indicate better overall sustainability performance. Similar approaches for composite sustainability assessment have been widely used in agricultural and environmental studies (53, 54).

3 Result and discussion

3.1 Descriptive statistics of soil properties

The descriptive statistics of soil physicochemical properties are presented in Table 1. Considerable variability was observed among the measured attributes, indicating heterogeneous soil fertility conditions across the rice-growing region of Telangana. Such variability provides the basis for precision agriculture because differences in nutrient status among locations can be exploited through site-specific nutrient management. Soil pH ranged from 6.62 to 7.82, with a mean value of 7.21. The low coefficient of variation (CV = 5.41%) indicates relatively uniform soil reaction throughout the study area. Similar low variability in pH has been reported in intensively cultivated soils of India by Moharana et al. (55) Ramulu et al. (56), and Nogiya et al. (39). The narrow pH range suggests that soil reaction contributes less to overall fertility differentiation than nutrient-related attributes. Electrical conductivity (EC) varied from 0.14 to 1.68 dS m^-¹, with a mean value of 0.78 dS m^-¹. Although all samples remained within the non-saline range, EC exhibited the highest variability among the measured properties (CV = 65.38%). This indicates the presence of substantial within-field heterogeneity and agrees with the findings of Jena et al. (10) and Shukla et al. (19), who reported pronounced EC variability under intensive agricultural systems. The greater variation in EC compared with pH suggests that localized differences in field conditions contribute to spatial heterogeneity within the study area. Organic carbon content ranged from 0.22 to 0.72%, with a mean value of 0.44%, indicating predominantly low to medium organic carbon status. The moderate variability (CV = 36.36%) demonstrates uneven distribution of soil organic matter across the landscape. Comparable OC variability has been reported in semi-arid agricultural regions by Zeraatpisheh et al. (15), Behera et al. (57), Selmy et al. (58) and Venugopal et al. (1). The observed variation in organic carbon is particularly important because it reflects differences in inherent fertility among sampling locations and indicates the existence of distinct fertility gradients within the study area. Available nitrogen ranged from 78 to 197 kg ha^-¹, with a mean value of 126.56 kg ha^-¹ and a CV of 26.68%, indicating moderate spatial variability. Available phosphorus (P₂O₅) varied between 60 and 165 kg ha^-¹ with a mean value of 111.12 kg ha^-¹, while available potassium (K₂O) ranged from 180 to 580 kg ha^-¹ with a mean value of 309.84 kg ha^-¹. Potassium exhibited greater variability (CV = 43.02%) than nitrogen and phosphorus, suggesting stronger spatial differences in K availability across the study area. Similar nutrient variability has been reported in precision agriculture studies by Behera et al. (9), Moharana et al. (18), Kumar et al. (59), Sharma and Sood (60) and Rajashekhar et al. (61). Based on the coefficient of variation, the degree of variability followed the order EC > K₂O > OC > P₂O₅ > N > pH. This pattern indicates that nutrient-related properties exhibited considerably greater heterogeneity than soil reaction. Similar observations have been reported by Verma et al. (7), Behera et al. (9), Reza et al. (62) and Moharana et al. (18), who identified nutrient variability as the primary basis for management zone delineation. The moderate to high variability observed for OC and available nutrients demonstrates the presence of distinct fertility gradients across the study area. These findings indicate that uniform fertilizer recommendations may not adequately address local nutrient requirements. Consequently, delineation of management zones based on soil fertility variability could provide a practical framework for site-specific nutrient management, enabling more efficient fertilizer use and improved productivity in rice-based production systems of northern Telangana.

3.2 Geostatistical analysis of soil properties

The semivariogram parameters of the measured soil properties are presented in Table 2. All variables exhibited clear spatial structure, indicating that soil variability within the study area was spatially organized rather than randomly distributed. The best-fitted experimental semivariogram models for individual soil properties are illustrated in Figure 6a–f. Following logarithmic transformation, spherical models provided the best fit for pH, EC, and organic carbon, whereas exponential models best described available N, P₂O₅, and K₂O (Table 2). Similar model behaviour has been reported in precision agriculture studies conducted under semi-arid agroecosystems, where relatively stable soil properties frequently follow spherical structures while nutrient-related variables are better represented by exponential models (1, 10, 60, 63, 64). The nugget-to-sill ratios ranged from 0 to 13.28%, indicating strong spatial dependence for all soil properties according to the classification proposed by Cambardella et al. (33). The consistently low nugget effects demonstrate high spatial continuity and suggest that the observed variability was systematic rather than random. Similar levels of spatial dependence have been reported in soil fertility zoning studies conducted in India and other intensively cultivated regions (8, 13, 18). The strong spatial structure observed in the present study confirms the suitability of these soil variables for kriging interpolation and management zone delineation.

Figure 6

Spatial ranges varied from 694 m for pH to 1592 m for available K₂O, indicating differences in the scale of spatial continuity among soil properties (Figures 6a-f). Nutrient-related variables generally exhibited larger ranges than pH and EC, suggesting broader patterns of nutrient distribution across the landscape. Comparable spatial ranges have been reported by Diacono et al. (21), Yuan et al. (65) and Nogiya et al. (39), who observed that nutrient attributes often show larger zones of influence than relatively stable soil chemical properties. The relatively large ranges obtained in the present study indicate that field variability can be effectively represented through management zones rather than uniform field-scale management. Cross-validation results showed satisfactory prediction accuracy, as reflected by the low RMSE values obtained for all variables. The combination of strong spatial dependence and acceptable prediction errors confirms the reliability of ordinary kriging for representing the spatial variability of soil fertility attributes. Similar interpolation performance has been reported by Moharana et al. (18) and Oppong Sarkodie et al. (66), who demonstrated that strong spatial dependence improves kriging accuracy in precision agriculture applications. The kriging maps (Figures 7a-f) revealed distinct spatial patterns among soil properties. Soil pH showed relatively uniform distribution throughout the study area, with only minor increases in the western and north-western regions. This agrees with the low coefficient of variation observed in the descriptive statistics and indicates that soil reaction contributes less to fertility differentiation than nutrient-related variables.

Figure 7

In contrast, organic carbon and available nutrients exhibited pronounced spatial variability. Higher OC concentrations were predominantly observed in the western and central portions of the study area, whereas comparatively lower values occurred in the south-eastern region. Available nitrogen displayed a similar distribution pattern, indicating strong agreement between OC and N spatial trends. Comparable relationships between organic carbon and nitrogen distribution have been reported by Behera et al. (9) and Moharana et al. (18), where both variables represented major contributors to soil fertility variability. Available phosphorus exhibited a contrasting distribution, with relatively higher concentrations occurring in the central and south-eastern parts of the study area. Similar phosphorus enrichment patterns have been reported in intensively cultivated rice-based systems by Rahul et al. (4) and Srinivasan et al. (8). In comparison, available potassium showed higher concentrations towards the western region, closely corresponding with areas having higher organic carbon and nitrogen levels. Similar K distribution patterns associated with fertility gradients have been reported by Kumar et al. (59) and Rajashekhar et al. (61). Collectively, the kriging maps revealed a distinct fertility gradient across the study area, with relatively higher fertility in the western and north-western regions and comparatively lower fertility in the south-eastern areas. The agreement between descriptive statistics and geostatistical analysis indicates that OC, N, P₂O₅, and K₂O are the principal contributors to spatial fertility variability. Similar findings have been reported by Behera et al. (9), Moharana et al. (18), Sharma and Sood (60) and Srinivasan et al. (8), who identified nutrient-related variables as the dominant factors controlling management zone formation.

From a precision agriculture perspective, the identified fertility gradients demonstrate that uniform fertilizer recommendations are unlikely to satisfy the heterogeneous nutrient requirements of the study area. Delineation of management zones based on these spatial patterns would enable fertilizer inputs to be adjusted according to local soil conditions, thereby improving nutrient-use efficiency, reducing unnecessary input application, and enhancing the sustainability of rice production in northern Telangana.

3.3 Correlation analysis

The Pearson correlation matrix among the measured soil properties is presented in Figure 8. Significant positive correlations (p < 0.01) were observed among most variables, indicating that the soil attributes exhibited coordinated spatial behaviour across the study area. The strong interrelationships among soil properties suggest that variability in soil fertility was governed by common underlying factors rather than independent processes. Soil pH showed positive associations with EC, organic carbon, and available nutrients; however, the relatively narrow pH range and low coefficient of variation indicate that its contribution to overall fertility variability was limited. Similar observations have been reported by Verma et al. (7) and Moharana et al. (18), who found that pH generally exhibits lower spatial variability than nutrient-related attributes in intensively cultivated agricultural systems. Electrical conductivity was positively correlated with organic carbon and available macronutrients, indicating that EC variability reflected differences in overall soil fertility rather than salinity constraints. This interpretation is consistent with the kriging maps, which showed non-saline conditions throughout the study area despite the relatively high variability in EC. Comparable relationships between EC and nutrient status have been reported by Jena et al. (10) and Srinivasan et al. (8) in precision agriculture studies. The strongest correlations were observed among organic carbon, available nitrogen, phosphorus, and potassium (r > 0.90, p < 0.01). The close association among these variables indicates the existence of a common fertility gradient across the study area. Similar strong relationships among OC and macronutrients have been reported by Verma et al. (7), Behera et al. (9), and Moharana et al. (18), where nutrient-related properties constituted the major sources of variability used for management zone delineation. However, the correlation coefficients observed in the present study were comparatively higher than those reported in many field-scale investigations, suggesting stronger covariance among fertility parameters under the relatively homogeneous management conditions prevailing in the study area. The strong positive relationships among N, P₂O₅, and K₂O further indicate that these nutrients followed similar spatial trends and were likely influenced by common management factors. Similar nutrient co-variation has been reported in soil fertility zoning studies conducted under rice-based production systems (9, 13). The overlapping spatial patterns observed in the kriging maps provide additional support for these relationships. The correlation structure also provides an important basis for multivariate analysis. The high covariance among OC, N, P₂O₅, and K₂O suggests that much of the information contained in the dataset is shared among variables and can therefore be represented by a smaller number of latent factors. Similar conclusions have been reported by Srinivasan et al. (8) and Bhagwan et al. (2), where strong inter-variable correlations facilitated dimensionality reduction and improved management zone delineation. From a precision agriculture perspective, the observed correlation pattern indicates that areas with higher organic carbon generally coincide with greater nutrient availability, whereas nutrient-deficient locations are characterized by lower fertility status. This coordinated variability provides a strong basis for identifying homogeneous management zones and implementing site-specific nutrient management strategies. Consequently, differential fertilizer application based on management zones could improve nutrient-use efficiency and reduce unnecessary fertilizer inputs in the rice-growing soils of Telangana.

Figure 8

3.4 Principal component analysis of soil properties

Principal component analysis (PCA) was performed to identify the dominant sources of soil variability and to reduce redundancy among the measured soil properties (67). The eigenvalues and component loadings are presented in Table 3, 4, while the PCA biplot and scree plot are shown in Figure 9–11. The first two principal components collectively explained 95.94% of the total variance, indicating that the multidimensional soil dataset could be effectively represented by two latent factors. Similar levels of variance explained have been reported in management zone studies where a limited number of components captured most of the variability associated with soil fertility (2, 8, 9). The first principal component (PC1) accounted for 87.25% of the total variance and exhibited strong positive loadings for organic carbon, available nitrogen, phosphorus, and potassium. The dominance of these variables indicates that PC1 represents the primary fertility gradient within the study area. The high contribution of OC and macronutrients is consistent with the correlation analysis, which revealed strong associations among these variables, and with the kriging maps, which showed similar spatial distribution patterns. Collectively, these results indicate that soil fertility variability in the study area is largely governed by coordinated changes in nutrient availability rather than by isolated variations in individual soil properties. Comparable studies conducted under precision agriculture frameworks have similarly identified organic carbon and macronutrients as the major contributors to management zone delineation. Behera et al. (9) and Moharana et al. (18) reported that nutrient-related variables dominated the first principal component and represented the most important indicators of soil fertility variability. However, the proportion of variance explained by PC1 in the present study was higher than that commonly reported in field-scale investigations, suggesting stronger covariance among fertility parameters and a more distinct fertility gradient within the rice-growing soils of Telangana. The second principal component (PC2) explained 8.69% of the total variance and was mainly influenced by pH and EC. Compared with PC1, the contribution of PC2 was relatively small, indicating that soil chemical conditions played a secondary role in explaining variability across the study area. Similar secondary components associated with soil reaction and chemical properties have been reported by Shahinzadeh et al. (68) and Salem et al. (13). The lower contribution of pH and EC further supports the results obtained from descriptive statistics and geostatistical analysis, which indicated limited variability in soil reaction compared with nutrient-related properties. The remaining components individually explained only a small proportion of the variance, demonstrating that most of the meaningful information contained in the dataset was captured by the first two components. Similar observations have been reported in multivariate studies where dimensionality reduction enabled efficient characterization of soil fertility patterns without substantial information loss (19, 66). The PCA biplot (Figure 9) showed close clustering of OC, N, P₂O₅, and K₂O vectors, confirming their strong positive association and supporting the existence of a common fertility gradient. In contrast, pH and EC occupied relatively distinct positions within the ordination space, indicating comparatively weaker contributions to overall variability. The distribution of sampling points along PC1 further demonstrated clear differences in fertility status among locations, corroborating the spatial patterns identified through kriging. The variable correlation circle (Figure 11) provided additional evidence that organic carbon and macronutrients were the principal drivers of soil variability. Similar clustering patterns have been reported by Behera et al. (9), Moharana et al. (18), and Srinivasan et al. (8), who identified these variables as the most suitable indicators for delineating management zones in precision agriculture systems. The strong agreement among PCA, correlation analysis, and geostatistical mapping demonstrates the robustness of the identified fertility gradients. From a practical perspective, the results suggest that variability in crop productivity within the study area is primarily associated with differences in organic matter and nutrient availability. Consequently, management interventions aimed at improving soil organic carbon and optimizing nutrient inputs are expected to have the greatest impact on productivity. Furthermore, the dominance of a limited number of fertility-related variables indicates that management zones can be delineated using a reduced set of indicators without significant loss of information. Such an approach simplifies soil monitoring and supports the implementation of site-specific nutrient management strategies. By aligning fertilizer inputs with local soil fertility conditions, management zones can improve nutrient-use efficiency, reduce unnecessary fertilizer application, lower production costs, and contribute to sustainable rice production in northern Telangana. Overall, the PCA results confirm that soil fertility variability in the study area is predominantly controlled by a common nutrient-driven gradient represented by organic carbon, available nitrogen, phosphorus, and potassium. These findings provide a strong scientific basis for subsequent management zone delineation and the development of site-specific nutrient management strategies for rice-based production systems.

Figure 9

Figure 10

Figure 11

3.5 Determination and characterization of management zones

3.5.1 Determination of management zones

To determine the optimal number of management zones, clustering validity indices were evaluated for different cluster numbers (9, 13, 69). The Partition Coefficient (PC), Fuzzy Performance Index (FPI), and Normalized Classification Entropy (NCE) were used to assess clustering performance (47, 70) (Figure 12). The PC showed a decreasing trend with an increase in the number of clusters, while both FPI and NCE attained minimum values at three clusters, indicating improved clustering efficiency and clearer group separation (71). In addition, the Xie–Beni (XB) index, which evaluates cluster compactness and separation, exhibited the lowest value at three clusters (72, 73) (Figure 13). These results collectively indicate that three management zones provide the most optimal management zones in study area (Figure 14).

Figure 12

Figure 13

Figure 14

To identify the optimal number of management zones (MZs), clustering validity indices were evaluated for 2–10 clusters. The Partition Coefficient (PC) decreased steadily with increasing clusters, while the Fuzzy Performance Index (FPI) and Normalized Classification Entropy (NCE) reached minimum values at three clusters. A higher PC coupled with lower FPI and NCE indicates better clustering efficiency and improved separation among groups. Similarly, the Xie–Beni (XB) index, which quantifies compactness within clusters and separation between clusters, was lowest for three clusters and increased when more than five clusters were considered. Collectively, these indices indicate that three management zones provide the most balanced solution, optimizing within-zone homogeneity and between-zone heterogeneity. These results are consistent with previous studies employing fuzzy clustering for site-specific management, where three to four zones typically captured soil and crop variability effectively (46, 59, 71, 74, 75). Management zones were delineated using ArcGIS, and their spatial distribution is presented in Figure 14.

3.5.2 Characteristics of delineated management zones

Soil properties exhibited clear stratification across the three delineated MZs (Table 5). Soil pH increased progressively from MZ-1 (7.07) to MZ-3 (7.64), with all zones differing significantly (p < 0.05). Electrical conductivity (EC) followed a similar trend, with MZ-3 showing the highest value (1.60 dS m^-¹), indicating greater soluble salt accumulation. Organic carbon content varied from 0.31% in MZ-1 to 0.69% in MZ-3, reflecting the spatial distribution of organic matter inputs. Corresponding variations were observed for macronutrients: nitrogen ranged from 99 kg ha^-¹ in MZ-1 to 182 kg ha^-¹ in MZ-3, while phosphorus (P₂O₅) and potassium (K₂O) increased from 83 to 159 kg ha^-¹ and 210 to 561 kg ha^-¹, respectively. The differences among zones were statistically significant, as confirmed by ANOVA tests (2, 8, 10, 18, 19, 76).

The observed heterogeneity underscores the importance of implementing site-specific nutrient management. Uniform fertilizer application could either under- or over-supply nutrients, reducing efficiency and potentially causing environmental losses (4, 9, 10, 77). Delineating management zones enables targeted interventions, improving nutrient-use efficiency and supporting sustainable crop production. The spatial distribution of the three MZs is illustrated in Figure 14, providing a practical framework for precision fertilization in the study area.

3.5.3 Validation, fertilizer recommendations and economic implications per management zone

3.5.3.1 Validation of management zones and fertilizer recommendations

The delineation of management zones (MZs) enabled the development of site-specific fertilizer recommendations using STCR-based targeted yield Equations 8–10, resulting in substantial reductions in fertilizer inputs compared with the conventional farmer fertilizer practice (Table 6). The farmer-applied fertilizer dose was 280 kg N ha^-¹, 80 kg P₂O₅ ha^-¹, and 120 kg K₂O ha^-¹. In contrast, the recommended fertilizer doses varied considerably among management zones according to their nutrient requirements, highlighting the spatial heterogeneity of soil fertility across the study area. Among the three management zones, MZ-1 required the highest fertilizer inputs (235 kg N ha^-¹, 28 kg P₂O₅ ha^-¹, and 103 kg K₂O ha^-¹), whereas MZ-3 required the lowest nutrient application rates (172 kg N ha^-¹, 0 kg P₂O₅ ha^-¹, and 60 kg K₂O ha^-¹). MZ-2 received intermediate fertilizer doses of 203 kg N ha^-¹, 0 kg P₂O₅ ha^-¹, and 88 kg K₂O ha^-¹. The complete omission of phosphorus fertilizer in MZ-2 and MZ-3 indicates that the indigenous soil phosphorus supply was sufficient to meet crop demand, eliminating the need for additional phosphorus fertilization. The observed variability in fertilizer recommendations among management zones reflects differences in soil nutrient availability, nutrient retention capacity, organic matter content, and overall soil productivity potential. According to the Soil Test Crop Response (STCR) approach, fertilizer requirements decrease as the contribution of nutrients from indigenous soil sources increases (78). Therefore, the lower fertilizer requirements observed in MZ-2 and MZ-3 suggest greater native nutrient-supplying capacity and improved soil fertility status compared with MZ-1. Similar observations were reported by Ramamoorthy et al. (79) and Bhagwan et al. (2), who demonstrated that soil test-based fertilizer recommendations provide more accurate nutrient prescriptions than blanket recommendations by accounting for the nutrient contribution from the soil.

The adoption of site-specific nutrient management (SSNM) resulted in considerable fertilizer savings across all management zones. Nitrogen savings ranged from 45 kg ha^-¹ in MZ-1 to 108 kg ha^-¹ in MZ-3, corresponding to reductions of 16.1% and 38.6%, respectively, relative to the farmer practice. Likewise, phosphorus savings ranged from 52 kg ha^-¹ in MZ-1 to 80 kg ha^-¹ in MZ-2 and MZ-3, representing reductions of 65–100% in phosphorus fertilizer use. Potassium savings varied from 17 kg ha^-¹ in MZ-1 to 60 kg ha^-¹ in MZ-3, equivalent to reductions ranging from 14.2% to 50.0%. The highest nutrient savings were recorded in MZ-3, indicating that conventional fertilizer practices may result in considerable nutrient over-application in areas with relatively high native fertility.

These findings are consistent with the principles of precision agriculture, which advocate matching nutrient application rates with spatially variable crop requirements. Mulla (80) reported that precision agriculture technologies can reduce fertilizer application rates by 10–40% while improving nutrient-use efficiency and maintaining crop productivity. Similarly, Gebbers and Adamchuk (81) and Dobermann et al. (82) emphasized that site-specific nutrient management improves resource-use efficiency, profitability, and environmental sustainability through optimized input application. The nutrient savings observed in the present study fall within or exceed the ranges reported in earlier precision agriculture investigations, highlighting the effectiveness of management-zone-based fertilizer recommendations. The elimination of phosphorus fertilizer in MZ-2 and MZ-3 is particularly important from both agronomic and environmental perspectives. Phosphorus is a finite resource and excessive phosphorus application often results in nutrient accumulation in soils, increasing the risk of eutrophication through runoff losses. Johnston and Bruulsema (83) emphasized that fertilizer recommendations should consider residual soil nutrient reserves and indigenous nutrient supply before prescribing external fertilizer inputs. Therefore, the omission of phosphorus fertilizer in these management zones demonstrates the practical value of soil-test-based nutrient management in reducing fertilizer costs while maintaining crop productivity. The observed fertilizer savings have direct economic implications because fertilizer expenditure constitutes a major component of cultivation costs. Reductions in fertilizer use not only lower production costs but also improve fertilizer-use efficiency and overall farm profitability. Furthermore, minimizing excessive nutrient application can reduce environmental risks associated with nitrate leaching, phosphorus runoff, and greenhouse gas emissions. Cassman et al. (84) emphasized that improving nutrient-use efficiency is essential for increasing agricultural productivity while minimizing environmental degradation. Therefore, management-zone-based fertilizer recommendations provide both economic and environmental benefits by enhancing nutrient stewardship and promoting sustainable agricultural intensification. Overall, compared with the conventional farmer fertilizer practice, management-zone-based nutrient management reduced fertilizer use by up to 38.6% N, 100% P₂O₅, and 50.0% K₂O. These findings demonstrate that integrating management-zone delineation with STCR-based fertilizer recommendations provides an effective framework for precision nutrient management, improved nutrient-use efficiency, reduced input costs, and environmentally sustainable crop production.

3.5.3.2 Economics of management zone-based nutrient management

The economic analysis revealed that management-zone-based fertilizer recommendations substantially improved grain yield, profitability, and benefit–cost (B:C) ratio compared with the conventional farmer fertilizer practice (Table 7). The economic gains achieved under site-specific nutrient management were closely associated with optimized fertilizer use and enhanced nutrient-use efficiency. By aligning fertilizer application rates with spatially variable soil fertility conditions, cultivation costs were reduced while crop productivity and economic returns increased.

Grain yield increased progressively from MZ-1 (59 q ha^-¹) to MZ-3 (67 q ha^-¹), whereas the farmer fertilizer practice recorded the lowest yield (53 q ha^-¹). The yield advantage under management-zone-based nutrient management ranged from 11.3% in MZ-1 to 26.4% in MZ-3 over the conventional practice. The higher yields obtained under site-specific nutrient management indicate improved synchronization between crop nutrient demand and nutrient supply, resulting in enhanced nutrient uptake and utilization efficiency. Similar improvements in crop productivity under site-specific nutrient management have been reported by Bhagwan et al. (85), who observed significant yield increases in maize due to balanced nutrient application based on soil fertility variability. The cost of cultivation was lower under all management zones than under the conventional farmer fertilizer practice. The highest cultivation cost was recorded under the farmer fertilizer dose (₹76,190.04 ha^-¹), whereas MZ-3 recorded the lowest cultivation cost (₹66,715.04 ha^-¹), followed by MZ-2 (₹68,820.04 ha^-¹) and MZ-1 (₹71,695.04 ha^-¹). The reduction in production costs was primarily attributed to substantial fertilizer savings achieved through site-specific nutrient recommendations. This finding confirms that blanket fertilizer application often leads to unnecessary expenditure on fertilizers without corresponding increases in productivity (86). Gross returns increased substantially under management-zone-based fertilization. The highest gross profit was obtained in MZ-3 (₹160,130 ha^-¹), followed by MZ-2 (₹148,180 ha^-¹) and MZ-1 (₹141,010 ha^-¹), whereas the farmer fertilizer practice generated only ₹126,670 ha^-¹. Higher gross returns were primarily associated with increased grain yield resulting from balanced nutrient management and improved nutrient-use efficiency. Similar results were reported by Bongiovanni and Lowenberg-DeBoer (87), who found that precision agriculture technologies improve economic returns through efficient resource allocation and optimized input use. Net profit followed a similar trend, with the highest net return recorded in MZ-3 (₹93,415 ha^-¹), followed by MZ-2 (₹79,360 ha^-¹) and MZ-1 (₹69,315 ha^-¹), compared with only ₹50,480 ha^-¹ under the conventional farmer fertilizer practice. Relative to the farmer practice, net profit increased by 37.3%, 57.2%, and 85.0% in MZ-1, MZ-2, and MZ-3, respectively. The substantial increase in profitability reflects the combined effect of reduced fertilizer expenditure and increased crop productivity under management-zone-based nutrient management. The benefit–cost ratio further confirmed the economic superiority of the management-zone approach. The B:C ratio increased from 1.66 under the conventional fertilizer practice to 1.97, 2.15, and 2.40 in MZ-1, MZ-2, and MZ-3, respectively. The highest B:C ratio observed in MZ-3 indicates that every rupee invested generated ₹2.40 in returns. These findings corroborate the observations of Robert (88) and Gebbers and Adamchuk (81), who reported that precision agriculture technologies improve economic efficiency by optimizing input use and increasing crop productivity. The superior economic performance of MZ-3 can be attributed to its relatively higher indigenous nutrient-supplying capacity, lower fertilizer requirement, reduced cultivation cost, and higher grain yield. The results indicate that management zones with better soil fertility status can sustain substantial reductions in fertilizer inputs without compromising productivity, thereby generating higher economic returns. Similar conclusions were reached by Kitchen et al. (89), who demonstrated that management-zone-based nutrient management effectively captures spatial variability and improves economic efficiency through variable-rate fertilizer application. Overall, compared with the conventional farmer fertilizer practice, management-zone-based nutrient management increased grain yield by up to 26.4%, enhanced net profit by 85.0%, improved the benefit–cost ratio from 1.66 to 2.40, and simultaneously reduced fertilizer consumption by up to 38.6% N, 100% P₂O₅, and 50.0% K₂O. These results clearly demonstrate that management-zone delineation combined with STCR-based fertilizer recommendations is an effective strategy for enhancing nutrient-use efficiency, improving farm profitability, reducing environmental risks, and promoting sustainable agricultural production under precision farming systems.

3.6 Sustainability indicators under different management zones

3.6.1 Sustainability indicators

The comparative assessment of sustainability indicators revealed distinct differences among the conventional farmer fertilizer practice and the management zone-based nutrient management strategies (Figure 15). The radar plot clearly demonstrated that management-zone-based fertilizer recommendations improved agronomic, economic, and resource-use efficiency indicators compared with the farmer practice. Among all treatments, MZ-3 consistently exhibited the highest normalized values for grain yield, net profit, benefit–cost (B:C) ratio, nitrogen saving, and potassium saving, indicating superior overall performance. Phosphorus saving was highest in both MZ-2 and MZ-3, where phosphorus fertilizer application was completely eliminated owing to sufficient indigenous soil phosphorus supply. In contrast, the farmer fertilizer practice recorded the lowest values for most sustainability indicators, reflecting inefficient resource utilization and lower economic returns. The progressive improvement in sustainability indicators from Farmer practice to MZ-3 highlights the effectiveness of management-zone-based nutrient management in optimizing fertilizer inputs according to site-specific soil fertility conditions. MZ-1 showed moderate improvements over the conventional practice, whereas MZ-2 and MZ-3 exhibited substantial gains in productivity, profitability, and nutrient-use efficiency. The balanced enhancement across all indicators in MZ-3 suggests a more effective synchronization between nutrient supply and crop demand, resulting in higher crop productivity with lower fertilizer inputs. These findings support the concept that precision nutrient management can simultaneously improve agronomic performance and resource-use efficiency by accounting for spatial variability in soil properties and nutrient availability. The superior performance of MZ-3 can be attributed to its relatively higher indigenous nutrient-supplying capacity, which enabled substantial reductions in fertilizer application without compromising crop productivity. The reductions in nitrogen, phosphorus, and potassium fertilizer requirements observed in MZ-3 indicate more efficient utilization of native soil fertility reserves. Similar observations have been reported by Doerge (90) and Fraisse et al. (91), who demonstrated that management zones effectively capture within-field variability and facilitate optimized input application. Likewise, Mulla (80) reported that precision agriculture technologies enhance nutrient-use efficiency and improve sustainability by reducing unnecessary fertilizer inputs while maintaining crop productivity. The improved economic indicators observed under MZ-2 and MZ-3 further emphasize the benefits of site-specific nutrient management. Higher net profit and B:C ratio were achieved through a combination of increased grain yield and reduced fertilizer expenditure (92). Bongiovanni and Lowenberg-DeBoer (87) reported that precision agriculture improves profitability primarily through optimized resource allocation and reduced input costs. Similarly, Gebbers and Adamchuk (81) highlighted that site-specific management practices increase both economic and environmental efficiency by matching agricultural inputs with local crop requirements. Therefore, the higher economic returns observed under MZ-3 demonstrate the practical value of management-zone-based nutrient management for improving farm profitability while reducing production costs. From an environmental perspective, the substantial nutrient savings achieved under management-zone-based nutrient management contribute directly to sustainable agricultural production. Reduced fertilizer application lowers the risk of nutrient losses through leaching, runoff, volatilization, and denitrification, thereby minimizing environmental pollution. Cassman et al. (84) emphasized that improving nutrient-use efficiency is essential for achieving sustainable agricultural intensification because it simultaneously enhances productivity and reduces environmental degradation. The complete elimination of phosphorus fertilizer in MZ-2 and MZ-3 is particularly noteworthy because excessive phosphorus application is often associated with nutrient accumulation in soils and eutrophication of surface water bodies. Consequently, the management-zone approach supports both economic and environmental sustainability through efficient nutrient stewardship.

Figure 15

3.6.2 Sustainability Performance Index

The Sustainability Performance Index (SPI) effectively integrated multiple agronomic, economic, and nutrient-use efficiency indicators into a single metric, providing a comprehensive assessment of overall sustainability performance among management zones (Figure 16). The SPI values clearly differentiated the treatments, with MZ-3 recording the highest value (1.000), followed by MZ-2 (0.690) and MZ-1 (0.440), whereas the farmer fertilizer practice recorded the lowest value (0.000). The progressive increase in SPI values from Farmer practice to MZ-3 demonstrates the substantial improvement in overall sustainability achieved through management-zone-based nutrient management. The highest SPI value observed in MZ-3 reflects its superior performance across all major sustainability dimensions, including productivity, profitability, and nutrient-use efficiency. The simultaneous achievement of higher grain yield, greater net profit, improved B:C ratio, and substantial fertilizer savings indicates a more balanced and sustainable production system (93). In contrast, the low SPI value associated with the conventional farmer fertilizer practice suggests lower resource-use efficiency and reduced economic and environmental sustainability. Similar integrated sustainability assessments have shown that precision agriculture and site-specific nutrient management technologies improve overall system sustainability by optimizing resource utilization and reducing production inefficiencies (80, 81). The results further indicate that management-zone-based nutrient management can serve as an effective decision-support strategy for achieving sustainable agricultural intensification. By integrating soil spatial variability into fertilizer recommendation frameworks, management zones facilitate more precise nutrient application, enhance nutrient-use efficiency, and improve farm profitability while reducing environmental risks. The markedly higher SPI value observed in MZ-3 demonstrates that areas with relatively favourable soil fertility status can achieve substantial improvements in sustainability when fertilizer inputs are optimized according to site-specific conditions. Overall, the radar plot and Sustainability Performance Index consistently identified MZ-3 as the most sustainable management zone, followed by MZ-2 and MZ-1, while the conventional farmer fertilizer practice exhibited the poorest sustainability performance. These findings demonstrate that integrating management-zone delineation with STCR-based nutrient management provides a robust framework for enhancing productivity, profitability, nutrient-use efficiency, and environmental sustainability. Therefore, management-zone-based nutrient management represents a promising approach for precision farming and climate-smart agricultural development (94).

Figure 16

4 Conclusions

The study clearly establishes that soils in the study area exhibit high spatial variability in pH, EC, organic carbon and macronutrient status. Conventional blanket fertilizer recommendations fail to account for this heterogeneity, resulting in nutrient imbalance and inefficient input use. By integrating geostatistical analysis, PCA and geographically weighted clustering techniques, the study successfully delineated three management zones with distinct soil fertility characteristics. The findings confirm that management-zone-based site-specific nutrient management can enhance nutrient-use efficiency, reduce nutrient losses and support sustainable rice production in semi-arid environments. This approach enables fertilizer application according to localized soil requirements, thereby minimizing both under- and over-application.

Furthermore, the methodology highlights the potential for integrating digital soil maps, GIS-based decision support tools, and precision agriculture technologies to further refine fertilizer recommendations. Future research should focus on validating these management zones through field experimentation, developing crop-specific fertilizer prescriptions using the STCR approach, and extending the methodology to other crops and agro-climatic regions. Overall, the study provides a robust scientific framework for advancing precision nutrient management and promoting environmentally sustainable rice production systems.

References

• 1

  VenugopalAKannanBGanapatiPSKrishnanAManikandanKMathiyazhaganVet al. Nutrient variability mapping and demarcating management zones by employing fuzzy clustering in southern coastal region of Tamil Nadu, India. Sustainability. (2024) 16:2095. doi: 10.3390/su16052095

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  BhagwanPVAnjaiahTRavaliCCharyDSZamaniATUllahSet al. Delineation and evaluation of management zones for site-specific nutrient management using a geostatistical and fuzzy C mean cluster approach. Sci Rep. (2025) 15:20991. doi: 10.1038/s41598-025-07283-0

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  SinghSKKumarMSharmaBKTarafdarJC. Depletion of organic carbon, phosphorus, and potassium stock under a pearl millet based cropping system in the arid region of India. Arid Land Res Manage. (2007) 21:119–31. doi: 10.1080/15324980701236101

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  RahulTKumarNABiswaranjanDMohammadSBanwariLPriyankaGet al. Assessing soil spatial variability and delineating site-specific management zones for a coastal saline land in eastern India. Arch Agron Soil Sci. (2019) 65:1775–87. doi: 10.1080/03650340.2019.1578345

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  KhanHFarooqueAAAcharyaBAbbasFEsauTJZamanQU. Delineation of management zones for site-specific information about soil fertility characteristics through proximal sensing of potato fields. Agronomy. (2020) 10:1854. doi: 10.3390/agronomy10121854

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  TripathiRNayakAKShahidMLalBGautamPRajaRet al. Delineation of soil management zones for a rice cultivated area in eastern India using fuzzy clustering. CATENA. (2015) 133:128–136. doi: 10.1016/j.catena.2015.05.009

  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  VermaRRManjunathBLSinghNPKumarAAsolkarTChavanVet al. Soil mapping and delineation of management zones in the Western Ghats of coastal India. Land Degrad Dev. (2018) 29:4313–22. doi: 10.1002/ldr.3183

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  SrinivasanRShashikumarBNSinghSK. Mapping of soil nutrient variability and delineating site-specific management zones using fuzzy clustering analysis in eastern coastal region, India. J Indian Soc Remote Sens. (2022) 50:533–47. doi: 10.1007/s12524-021-01473-9

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  BeheraSKMathurRKShuklaAKSureshKPrakashC. Spatial variability of soil properties and delineation of soil management zones of oil palm plantations grown in a hot and humid tropical region of southern India. Catena. (2018) 165:251–9. doi: 10.1016/j.catena.2018.02.008

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  JenaRKBandyopadhyaySPradhanUKMoharanaPCKumarNSharmaGKet al. Geospatial modelling for delineation of crop management zones using local terrain attributes and soil properties. Remote Sens. (2022) 14:2101. doi: 10.3390/rs14092101

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 11

  SoropaGNyamangaraJNyakatawaEZNyapwereNLarkRM. Spatial variability of soil secondary and micronutrients under smallholder maize production system in sub-humid condition of Zimbabwe. Geoderma Regional. (2023) 34:e00693. doi: 10.1016/j.geodrs.2023.e00693

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  SperanzaEANaimeJDMVazCMPSantosJCFDInamasuRYLopesIDONet al. Delineating management zones with different yield potentials in soybean–corn and soybean–cotton production systems. Agriengineering. (2023) 5:1481–97. doi: 10.3390/agriengineering5030092

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  SalemHMSchottLRPiaskowskiJChapagainAYostJLBrooksEet al. Evaluating intra-field spatial variability for nutrient management zone delineation through geospatial techniques and multivariate analysis. Sustainability. (2024) 16:645. doi: 10.3390/su16020645

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  RamzanSNazirSAshrafIWaniMAWaniZMShafiqMU. Management zone delineation and spatial distribution of micronutrients in cold-arid region of India. Environ Monit Assess. (2021) 193:433. doi: 10.1007/s10661-021-09216-6

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  ZeraatpishehMBottegaELBakhshandehEOwliaieHRTaghizadeh-MehrjardiRKerryRet al. Spatial variability of soil quality within management zones: homogeneity and purity of delineated zones. Catena. (2022) 209:105835. doi: 10.1016/j.catena.2021.105835

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 16

  WangNXuDXueJZhangXHongYPengJet al. Delineation and optimization of cotton farmland management zone based on time series of soil-crop properties at landscape scale in south Xinjiang, China. Soil Tillage Res. (2023) 231:105744. doi: 10.1016/j.still.2023.105744

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 17

  ChatterjeeSSantraPMajumdarKGhoshDDasISanyalSK. Geostatistical approach for management of soil nutrients with special emphasis on different forms of potassium considering their spatial variation in intensive cropping system of West Bengal, India. Environ Monit Assess. (2015) 187:183. doi: 10.1007/s10661-015-4414-9

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 18

  MoharanaPCJenaRKPradhanUKNogiyaMTailorBLSinghRSet al. Geostatistical and fuzzy clustering approach for delineation of site-specific management zones and yield-limiting factors in irrigated hot arid environment of India. Precis Agric. (2020) 21:426–48. doi: 10.1007/s11119-019-09671-9

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  ShuklaAKBeheraSKBasumataryASarangthemIMishraRDuttaSet al. PCA and fuzzy clustering-based delineation of soil nutrient (S, B, Zn, Mn, Fe, and Cu) management zones of sub-tropical Northeastern India for precision nutrient management. J Environ Manage. (2024) 365:121511. doi: 10.1016/j.jenvman.2024.121511

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 20

  BelayAMSelassieYGTsegayeEAMeshaesheDTAddisHK. Spatial analysis of some soil chemical properties of the Amhara region in Ethiopia. Arab J Geosci. (2024) 17:205. doi: 10.1007/s12517-024-12003-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  DiaconoMDe BenedettoDCastrignanòARubinoPVittiCAbdelrahmanHMet al. A combined approach of geostatistics and geographical clustering for delineating homogeneous zones in a durum wheat field in organic farming. NJAS Wageningen J Life Sci. (2013) 64–65:47–57. doi: 10.1016/j.njas.2013.03.001

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  PanjalaPGummaMKMesapamS. Geospatial assessment of cropping pattern shifts and their impact on water demand in the Kaleshwaram lift irrigation project command area, Telangana. Front Remote Sens. (2024) 5:1451594. doi: 10.3389/frsen.2024.1451594

  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  ChanapathiTThatikondaSKeesaraVRPonguruNS. Assessment of water resources and crop yield under future climate scenarios: A case study in a Warangal district of Telangana, India. J Earth Syst Sci. (2020) 129:20. doi: 10.1007/s12040-019-1294-3

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  BhattacharyyaTChandranPRaySKPalDK. Soil classification following the US taxonomy: An Indian commentary. Soil Horiz. (2015) 56:1–16. doi: 10.2136/sh14-08-0011

  • CrossRef
  • Google Scholar
  • View reference in article

• 25

  MylavarapuSKNaiduMVSKavithaPReddyMS. Characterization, classification and evaluation of soils in semi-arid region of Mahanandi Mandal in Kurnool District of Andhra Pradesh. Jour Indian Socie Soil Scie. (2019) 67:125. doi: 10.5958/0974-0228.2019.00014.8

  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  OliverMAWebsterR. A tutorial guide to geostatistics: Computing and modelling variograms and kriging. Catena. (2014) 113:56–69. doi: 10.1016/j.catena.2013.09.006

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 27

  WalkleyABlackIA. An examination of the degtjareff method for determining soil organic matter, and a proposed modification of the chromic acid titration method. Soil Sci. (1934) 37:29–38. doi: 10.1097/00010694-193401000-00003

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 28

  SubbiahBVAsijaGL. A rapid procedure for the estimation of available nitrogen in soils. Curr Scienc. (1956) 25:259–60.

  • Google Scholar
  • View reference in article

• 29

  OlsenSRColeCVWatanabeFSDeanLA. Estimation of Available Phosphorus in Soils by Extraction with Sodium Bicarbonate. Circular No. 939. (1954). Washington, DC: U.S. Department of Agriculture.

  • Google Scholar
  • View reference in article

• 30

  HanwayJJHeidalH. Soil analysis methods as used in Iowa state college soil testing laboratory. Iowa State Coll Agric Bull. (1952) 57:1–31.

  • Google Scholar
  • View reference in article

• 31

  AliyuKTKamaraYAJibrinMJHuisingEJShehuMBAdewopoJBet al. Delineation of soil fertility management zones for site-specific nutrient management in the maize belt region of Nigeria. Sustainability. (2020) 12:9010. doi: 10.3390/su12219010

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 32

  De CairesSAMartinCSAtwellMAKayaFWuddiviraGAWuddiviraMN. Advancing soil mapping and management using geostatistics and integrated machine learning and remote sensing techniques: A synoptic review. Discover Soil. (2025) 2(1):53.

  • Google Scholar
  • View reference in article

• 33

  CambardellaCAMoormanTBNovakJMParkinTBKarlenDLTurcoRFet al. Field‐scale variability of soil properties in central Iowa soils. Soil Sci Soc Amer J. (1994) 58:1501–11. doi: 10.2136/sssaj1994.03615995005800050033x

  • CrossRef
  • Google Scholar
  • View reference in article

• 34

  GökmenVSürücüABudakMBilgiliAV. Modeling and mapping the spatial variability of soil micronutrients in the Tigris basin. J King Saud Univ Sci. (2023) 35:102724. doi: 10.1016/j.jksus.2023.102724

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 35

  ConceiçãoLASilvaLValeroCLouresLMaçãsB. Delineation of soil management zones and validation through the vigour of a fodder crop. AgriEngineering. (2024) 6:205–27. doi: 10.3390/agriengineering6010013

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 36

  JolliffeITCadimaJ. Principal component analysis: a review and recent developments. Philos Trans A Math Phys Eng Sci. (2016) 374(2065):20150202. doi: 10.1098/rsta.2015.0202

  • CrossRef
  • Google Scholar
  • View reference in article

• 37

  OuazaaSJaramillo-BarriosCIChaaliNAmayaYMQCarvajalJECRamosOM. Towards site specific management zones delineation in rotational cropping system: Application of multivariate spatial clustering model based on soil properties. Geoderma Regional. (2022) 30:e00564. doi: 10.1016/j.geodrs.2022.e00564

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 38

  Peter-JeromeHAdewopoJBKamaraAYAliyuKTDawakiMU. Assessing the spatial variability of soil properties to delineate nutrient management zones in smallholder maize-based system of Nigeria. Appl Environ Soil Sci. (2022) 2022:1–14. doi: 10.1155/2022/5111635

  • CrossRef
  • Google Scholar
  • View reference in article

• 39

  NogiyaMMoharanaPCMeenaRYadavBJangirAMalavLCet al. Spatial variability of soil variables using geostatistical approaches in the hot arid region of India. Environ Earth Sci. (2024) 83:432. doi: 10.1007/s12665-024-11737-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 40

  BrunsdonCFotheringhamASCharltonME. Geographically weighted regression: A method for exploring spatial nonstationarity. Geogr Anal. (1996) 28:281–98. doi: 10.1111/j.1538-4632.1996.tb00936.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 41

  PanXWangLHuangCWangSChenH. A novel weighted fuzzy c-means based on feature weight learning. Ifs. (2021) 41:6149–67. doi: 10.3233/JIFS-202779

  • CrossRef
  • Google Scholar
  • View reference in article

• 42

  LiYShiZLiFLiH-Y. Delineation of site-specific management zones using fuzzy clustering analysis in a coastal saline land. Comput Electron Agric. (2007) 56:174–86. doi: 10.1016/j.compag.2007.01.013

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 43

  TagarakisALiakosVFountasSKoundourasSGemtosTA. Management zones delineation using fuzzy clustering techniques in grapevines. Precis Agric. (2013) 14:18–39. doi: 10.1007/s11119-012-9275-4

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 44

  SunJVázquez ArellanoMWangYMouazenAM. Optimizing management zone delineation technique for high-dimensional and large-volume datasets in precision agriculture. In: StaffordJV, editor. Precision Agriculture 25 (Set Two Volumes). Leiden: Brill | Wageningen Academic (2025). p. 38–45. doi: 10.1163/9789004725232_003

  • CrossRef
  • Google Scholar
  • View reference in article

• 45

  BougiouklisJ-NBarouchasPEPetropoulosPTsesmelisDEMoustakasN. Precision soil sampling strategy for the delineation of management zones in olive cultivation using unsupervised machine learning methods. Sci Rep. (2025) 15:8253. doi: 10.1038/s41598-025-89395-1

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 46

  EnejiAEOtieVIsongAAkiEAkpanJAdamuMet al. Delineation of field management zones based on soil properties in the Sudan and Sahel savannas of northern Nigeria. J Agric Food Res. (2025) 21:101812. doi: 10.1016/j.jafr.2025.101812

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 47

  KeshavarziADe CairesSASintimHYKayaFKusiNYOGyasi-AgyeiYet al. Spatial variability and management zones: Leveraging geostatistics and fuzzy clustering. J Soil Sci Plant Nutr. (2025) 25:3633–51. doi: 10.1007/s42729-025-02357-4

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 48

  VermaRRSrivastavaTKSinghPManjunathBLKumarA. Spatial mapping of soil properties in Konkan region of India experiencing anthropogenic onslaught. PLoS ONE. (2021) 16:e0247177. doi: 10.1371/journal.pone.0247177

  • CrossRef
  • Google Scholar
  • View reference in article

• 49

  GomezKAGomezAA. Statistical Procedures for Agricultural Research. New York, NY, United States: John Wiley & Sons (1984). 2nd Edn.

  • Google Scholar
  • View reference in article

• 50

  CIMMYT. From Agronomic Data to Farmer Recommendations: An Economics Training Manual. Mexico, D.F. Mexico: International Maize and Wheat Improvement Center (CIMMYT (1988). Completely Revised Edition.

  • Google Scholar
  • View reference in article

• 51

  FAO. The Future of Food and Agriculture: Trends and Challenges. Rome: Food and Agriculture Organization of the United Nations (2017). Available online at: https://www.fao.org/3/i6583e/i6583e.pdf (Accessed June 30, 2026).

  • Google Scholar
  • View reference in article

• 52

  PrettyJ. Agricultural sustainability: concepts, principles and evidence. Philos Trans R Soc B Biol Sci. (2008) 363:447–65. doi: 10.1098/rstb.2007.2163

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 53

  NardoMSaisanaMSaltelliATarantolaS. Handbook on Constructing Composite Indicators: Methodology and User Guide. Paris, France: OECD Publishing (2005). doi: 10.1787/533411815016

  • CrossRef
  • Google Scholar
  • View reference in article

• 54

  SinghRKMurtyHRGuptaSKDikshitAK. An overview of sustainability assessment methodologies. Ecol Indic. (2009) 9:189–212. doi: 10.1016/j.ecolind.2008.05.011

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 55

  MoharanaPCJenaRKYadavB. Digital mapping of soil health card parameters and nutrient management zones in the Thar Desert regions of India using quantile regression forest techniques. Arab J Geosci. (2023) 16:560. doi: 10.1007/s12517-023-11670-0

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 56

  RamuluCHarikrishnaBReddyRPUma ReddyR. Identification of soil fertility constraints of erstwhile Warangal district, Telangana, India, using GIS for their precise management for sustainable crop productivity. Int J Plant Soil Sci. (2024) 36:826–36. doi: 10.9734/ijpss/2024/v36i64690

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 57

  BeheraSKShuklaAKPatraAKPrakashCTripathiAKumar ChaudhariSet al. Assessing farm-scale spatial variability of soil nutrients in central India for site-specific nutrient management. Arab J Geosci. (2022) 15:848. doi: 10.1007/s12517-022-10138-x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 58

  SelmySEl-AzizSAEl-DesokyAEl-SayedM. Characterizing, predicting, and mapping of soil spatial variability in Gharb El-Mawhoub area of Dakhla Oasis using geostatistics and GIS approaches. J Saudi Soc Agric Sci. (2022) 21:383–96. doi: 10.1016/j.jssas.2021.10.013

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 59

  KumarPSharmaMButailNPShuklaAKKumarP. Spatial variability of soil properties and delineation of management zones for Suketi basin, Himachal Himalaya, India. Environ Dev Sustain. (2023) 26:14113–38. doi: 10.1007/s10668-023-03181-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 60

  SharmaRSoodK. Characterization of spatial variability of soil parameters in apple orchards of Himalayan region using geostatistical analysis. Commun Soil Sci Plant Anal. (2020) 51:1065–77. doi: 10.1080/00103624.2020.1744637

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 61

  RajashekharDMadhaviABabuPSRamprakashTVaniKPCharyDS. Soil fertility status of paddy (Oryza sativa L.) growing soils of different agro-climatic zones, Telangana, India. Ajsspn. (2024) 10:616–30. doi: 10.9734/ajsspn/2024/v10i3375

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 62

  RezaSKNayakDCMukhopadhyaySChattopadhyayTSinghSK. Characterizing spatial variability of soil properties in alluvial soils of India using geostatistics and geographical information system. Arch Agron Soil Sci. (2017) 63:1489–98. doi: 10.1080/03650340.2017.1296134

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 63

  KumariMKumarRPadbhushanRSinghYKVimalBKKumariRet al. Investigating the spatial variability of soil parameters and mineralogical characterization in the tea growing area of Kishanganj district Bihar India. Sci Rep. (2025) 15:29372. doi: 10.1038/s41598-025-04561-9

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 64

  YenenehNEliasEFeyisaGL. Assessment of the spatial variability of selected soil chemical properties using geostatistical analysis in the north-western highlands of Ethiopia. Acta Agriculturae Scandinavica Section B — Soil Plant Sci. (2022) 72:1009–19. doi: 10.1080/09064710.2022.2142658

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 65

  YuanYMiaoYYuanFAta-UI-KarimSTLiuXTianYet al. Delineating soil nutrient management zones based on optimal sampling interval in medium- and small-scale intensive farming systems. Precis Agric. (2022) 23:538–58. doi: 10.1007/s11119-021-09848-1

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 66

  Oppong SarkodieVVašátRNěmečekKŠrámekVFadrhonsováVNeudertová HellebrandováKet al. Assessment of multivariate associations and spatial variability of forest soil properties and their stand factors in the Czech Republic. Soil Water Res. (2025) 20:32–42. doi: 10.17221/114/2024-SWR

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 67

  SharmaS. Applied Multivariate Techniques. Hoboken: Wiley (2008). Nachdr.

  • Google Scholar

• 68

  ShahinzadehNBabaeinejadTMohsenifarKGhanavatiN. Spatial variability of soil properties determined by the interpolation methods in the agricultural lands. Model Earth Syst Environ. (2022) 8:4897–907. doi: 10.1007/s40808-022-01402-w

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 69

  Muniammal VediappanDGodiAGollaB. Geospatial mapping of soil properties of forest types using the k-means fuzzy clustering approach to delineate site-specific nutrient management zones in Goa, India. J Indian Soc Remote Sens. (2025) 53:2057–85. doi: 10.1007/s12524-024-02082-y

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 70

  MetwallyMSShaddadSMLiuMYaoR-JAbdoAILiPet al. Soil properties spatial variability and delineation of site-specific management zones based on soil fertility using fuzzy clustering in a hilly field in Jianyang, Sichuan, China. Sustainability. (2019) 11:7084. doi: 10.3390/su11247084

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 71

  VenkateswarluMRallapalliSSinghAChalapathiGSSKumarSKatpatalYBet al. Macro and micronutrient based soil fertility zonation using fuzzy logic and geospatial techniques. Sci Rep. (2025) 15:26772. doi: 10.1038/s41598-025-12184-3

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 72

  XieXLBeniG. A validity measure for fuzzy clustering. IEEE Trans Pattern Anal Mach Intell. (1991) 13:841–7. doi: 10.1109/34.85677

  • CrossRef
  • Google Scholar

• 73

  MotaVCDamascenoFALeiteDF. Fuzzy clustering and fuzzy validity measures for knowledge discovery and decision making in agricultural engineering. Comput Electron Agric. (2018) 150:118–24. doi: 10.1016/j.compag.2018.04.011

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 74

  BarmanASheoranPYadavRKAbhishekRSharmaRPrajapatKet al. Soil spatial variability characterization: Delineating index-based management zones in salt-affected agroecosystem of India. J Environ Manage. (2021) 296:113243. doi: 10.1016/j.jenvman.2021.113243

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 75

  Arévalo-HernándezJJOliveiraEMDFerrazGAESPolanía-MontielDCLiscano SolanoALSilvaMLN. The delineation of management zones using soil quality indices for the cultivation of irrigated rice (Oryza sativa L.) in Huila, Colombia. Geoderma Regional. (2024) 39:e00886. doi: 10.1016/j.geodrs.2024.e00886

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 76

  ShashikumarBNKumarSGeorgeKJSinghAK. Soil variability mapping and delineation of site-specific management zones using fuzzy clustering analysis in a Mid-Himalayan watershed, India. Environ Dev Sustain. (2023) 25:8539–59. doi: 10.1007/s10668-022-02411-6

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 77

  ShuklaAKSinhaNKTiwariPKPrakashCBeheraSKLenkaNKet al. Spatial distribution and management zones for sulphur and micronutrients in Shiwalik Himalayan region of India. Land Degrad Dev. (2017) 28:959–69. doi: 10.1002/ldr.2673

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 78

  NagarajuBKadappaBPRangaiahKMKasturappaGHussainSVeerabadraiahMSet al. Soil test crop response based fertilizer calibration and soil nutrient forecasting for aerobic rice on Alfisols of Southern India. Sci Rep. (2026) 16:17252. doi: 10.1038/s41598-026-48172-4

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 79

  RamamoorthyBNarasimhamRLDineshRS. Fertilizer application for specific yield targets on Sonora 64 (wheat). Indian Farming. (1967) 17(5):43–45.

  • Google Scholar

• 80

  MullaDJ. Twenty five years of remote sensing in precision agriculture: Key advances and remaining knowledge gaps. Biosyst Eng. (2013) 114:358–71. doi: 10.1016/j.biosystemseng.2012.08.009

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 81

  GebbersRAdamchukVI. Precision agriculture and food security. Science. (2010) 327:828–31. doi: 10.1126/science.1183899

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 82

  DobermannAWittCDaweDAbdulrachmanSGinesHCNagarajanRet al. Site-specific nutrient management for intensive rice cropping systems in Asia. Field Crops Res. (2002) 74:37–66. doi: 10.1016/S0378-4290(01)00197-6

  • CrossRef
  • Google Scholar

• 83

  JohnstonAMBruulsemaTW. 4R nutrient stewardship for improved nutrient use efficiency. Proc Eng. (2014) 83:365–70. doi: 10.1016/j.proeng.2014.09.029

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 84

  CassmanKGDobermannAWaltersDTYangH. Meeting cereal demand while protecting natural resources and improving environmental quality. Annu Rev Environ Resour. (2003) 28:315–58. doi: 10.1146/annurev.energy.28.040202.122858

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 85

  BhagwanPVAnjaiahTRavaliCDeviMUNeelimaTLCharyDSet al. Delineating soil fertility management zones using geostatistics and fuzzy clustering in semi-arid maize systems in India. Environ Monit Assess. (2025) 197:1230. doi: 10.1007/s10661-025-14608-z

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 86

  Krishna MurthyRBasavarajaPKBhavyaNDeyPMohamed SaqueebullaHGangamurthaGVet al. Development and validation of soil test based fertilizer prescription equations for enhancing yield, uptake and nutrient use efficiency of foxtail millet (Setaria italica) under dryland condition. J Plant Nutr. (2023) 46:3082–100. doi: 10.1080/01904167.2023.2188065

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 87

  BongiovanniRLowenberg-DeboerJ. Precision agriculture and sustainability. Precis Agric. (2004) 5:359–87. doi: 10.1023/B:PRAG.0000040806.39604.aa

  • CrossRef
  • Google Scholar

• 88

  RobertPC. Precision agriculture: a challenge for crop nutrition management. Plant Soil. (2002) 247:143–9. doi: 10.1023/A:1021171514148

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 89

  KitchenNRSudduthKAMyersBDDrummondSTHongSY. Delineating productivity zones on claypan soil fields using apparent soil electrical conductivity. Comput Electron Agric. (2005) 46:285–308. doi: 10.1016/j.compag.2004.11.012

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 90

  DoergeTA. Management zone concepts. In: Site-Specific Management Guidelines. (Norcross, GA: Potash & Phosphate Institute). (1999).

  • Google Scholar

• 91

  FraisseCWSudduthKAKitchenNR. Delineation of site-specific management zones by unsupervised classification of topographic attributes and soil electrical conductivity. Transactions of the ASAE. (2001) 44(1):155–166. doi: 10.13031/2013.2296

  • CrossRef
  • Google Scholar

• 92

  MonzonJPCalviñoPASadrasVOZubiaurreJBAndradeFH. Precision agriculture based on crop physiological principles improves whole-farm yield and profit: A case study. Eur J Agron. (2018) 99:62–71. doi: 10.1016/j.eja.2018.06.011

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 93

  LiYShiZWuCLiHLiF. Determination of potential management zones from soil electrical conductivity, yield and crop data. J Zhejiang Univ Sci B. (2008) 9(1):68–76. doi: 10.1631/jzus.B071379

  • CrossRef
  • Google Scholar

• 94

  BeheraAKBeheraSKBasumataryASarangthemIMishraRDuttaSet al. PCA and fuzzy clustering-based delineation of soil nutrient (S, B, Zn, Mn, Fe, and Cu) management zones of sub-tropical Northeastern India for precision nutrient management. J Environ Manage. (2024) 365:121511. doi: 10.1016/j.jenvman.2024.121511

  • Pubmed Abstract
  • CrossRef
  • Google Scholar