Terrain-Integrated Soil Mapping Units (SMUs) for Precision Nutrient Management: A Case Study from Semi-Arid Tropics of India

Abstract

This study presents a terrain-integrated Soil Management Unit (SMU) framework for precision agriculture in semi-arid tropical basaltic soils. Using high resolution (10-ha grid) sampling across 4627 geo-referenced locations and machine learning-enhanced integration of terrain attributes with legacy soil maps, and (3) quantitative validation of intra-SMU homogeneity, 15 SMUs were delineated based on landform, soil depth, texture, and slope. Principal Component Analysis (PCA) revealed SMU11 as the most heterogeneous (68.8%). Geo-statistical analysis revealed structured variability in soil pH (range = 1173 m) and nutrients availability with micronutrient sufficiency following Mn > Fe > Cu > Zn, (Zn deficient in SMU13). Organic carbon strongly correlated with key nutrients (AvK, r = 0.83 and Zn, r = 0.86). This represents the first systematic implementation of terrain-integrated SMU delineation in India’s basaltic landscapes, demonstrating a potential for 20–25% input savings. The spatially explicit fertility-integrated SMU framework provides a robust basis for developing decision support systems aimed at optimizing location-specific nutrient and land management strategies.

1. Introduction

Global agricultural systems face unprecedented challenges in the 21st century, with soil degradation and climate change threatening the very foundation of food security [1]. The Food and Agriculture Organization estimates that nearly 33% of the world’s soils are already degraded, with this figure rising to 40% in developing countries like India [2]. In this context, the need for precision soil management has never been more urgent, particularly in semi-arid tropical regions where fragile ecosystems and intensive agriculture coexist [3]. Traditional agricultural practices, characterized by uniform input application across heterogeneous fields, have proven increasingly unsustainable [4]. Such blanket approaches not only lead to economic inefficiencies but also contribute to environmental degradation through nutrient leaching and greenhouse gas emissions [5]. Precision agriculture has emerged as a potential solution, with its core philosophy of “doing the right thing, in the right place, at the right time” [6]. However, the successful implementation of precision farming depends fundamentally on understanding and managing spatial variability in soil properties [7].

Soil variability is shaped by the complex interplay of five key factors: climate, organisms, relief, parent material, and time [8]. Among these, relief or terrain attributes (elevation, slope, aspect, and landform position) exert a particularly strong influence on soil formation and nutrient distribution [9]. For instance, summit positions typically show well-drained, nutrient-poor soils due to erosional processes, while foot slope and depression areas accumulate both sediments and nutrients [10]. These terrain–soil relationships are especially pronounced in tropical regions with intense rainfall patterns, where a single monsoon season can redistribute significant amounts of soil and nutrients [11].

The concept of Soil Management Units (SMUs) has evolved as a practical framework to capture this spatial variability [12]. SMUs represent homogeneous zones that respond similarly to management interventions, bridging the gap between detailed soil surveys and practical farm management [13]. Modern SMU delineation increasingly relies on digital soil mapping techniques that integrate remote sensing, terrain analysis, and geostatistics [14]. These approaches have demonstrated particular value in developing countries where traditional soil survey resources are limited [15]. In the Indian context, the semi-arid tropics present unique challenges for soil management. The Deccan Plateau’s basaltic soils, while inherently fertile, are prone to degradation under intensive cotton cultivation [16]. Previous attempts at SMU delineation in the region, such as Bhaskar (2015) [17], relied solely on soil physicochemical properties without incorporating critical terrain parameters. This limitation resulted in management units that failed to fully capture the landscape’s inherent variability, particularly in terms of hydrological processes and erosion patterns [18]. Recent advances in geospatial technologies have opened new possibilities for more robust SMU development. High-resolution digital elevation models (DEMs), coupled with machine learning algorithms, now allow for precise characterization of terrain attributes at farm scales [19]. Simultaneously, the proliferation of proximal soil sensors has made high-density soil sampling more accessible [20]. These technological developments create opportunities to revisit traditional SMU approaches and develop more effective precision farming strategies for smallholder systems [21].

This study builds upon these advances to address critical gaps in precision agriculture implementation for semi-arid tropical regions. We focus specifically on: (a) developing a comprehensive SMU framework that integrates both terrain parameters and soil properties; (b) quantifying the spatial variability of key soil fertility indicators across different SMUs; (c) evaluating the agronomic and economic implications of SMU-based management; (d) providing practical recommendations for smallholder farmers in resource-constrained environments. Our work contributes to the growing body of research on sustainable intensification in tropical agriculture [22], while addressing the specific challenges of India’s cotton-growing regions. The findings have relevance not only for local farmers but also for agricultural extension systems and policy-makers working to promote sustainable land management practices. The study area in Maharashtra’s Yavatmal district represents a typical semi-arid tropical landscape where declining soil health threatens the livelihoods of millions of smallholder farmers [23]. By developing and validating a robust SMU framework for this region, we aim to provide a model that can be adapted to similar environments across the global tropics. Our approach combines cutting-edge geospatial technologies with practical agronomic insights, creating a bridge between scientific innovation and on-farm implementation.

2. Materials and Methods

2.1. Study Area

The soil sampling was carried out in the Yavatmal district situated between the latitude of 19°47′30″ to 20°15′22″ N and the longitude of 78°24′10″ to 78°41′49″ E in central India (Figure 1). The area experiences hot dry summers and mild, dry winters and is generally dry throughout the year except during the peak of the southwest monsoon (July to August). The region has 120 to 150 days of growing period with an ustic soil moisture regime and hyperthermic soil temperature regime [24]. The average annual temperature is 26.8 °C and the mean annual precipitation is 911 mm [25]. The shallow to deep, reddish brown, clay to clay loam soils are reported in this region. Major crops grown in the area are cotton (Gossypium hirsutum), soybean (Glycine max), pigeon pea (Cajanus cajan), sorghum (Sorghum bicolor) in the rainy season and wheat (Triticum aestivum), as well as chickpea (Cicer arietinum) in the winter season.

2.2. Soil Mapping Units

The delineation of Soil Management Units (SMUs) was conducted through an integrated analysis of terrain and soil characteristics using established digital soil mapping methodologies [26] (Figure 2). Landform classification and slope mapping were derived from 30 m resolution Shuttle Radar Topography Mission (SRTM) DEM data [27], following terrain analysis protocols [9]. Six landform units were identified: alluvial plain, pediment, plateau, pediplain, undulating plain, and valleys, consistent with classifications used in similar semi-arid regions [16]. Slope gradients were categorized into two classes (1–3% as very gentle; 3–5% as moderate) based on optimal ranges for agricultural management [10]. Soil depth and texture data were obtained from 1:50,000 scale legacy soil maps [28], with depth classified into six categories (very shallow to very deep) and texture into three classes (clayey, clay loam, silty clay). These classification schemes have demonstrated 85–90% accuracy in predicting soil properties in tropical regions [16]. The integration of datasets in ArcGIS followed the weighted overlay approach [29], where each parameter (landform, slope, depth, texture) was assigned relative importance based on local soil–landscape relationships (

Table 1. Criteria Weights and Rationale for SMU delineations.

Parameter	Weight	Rationale	Reference
Landform	0.40	Controls hydrology	[10]
Slope	0.30	Affects erosion	[29]
Depth	0.20	Rooting limitation	[30]
Texture	0.10	Modifies processes	[16]

). This method has been shown to reduce within-unit variability by 25–30% compared to single-parameter classifications [30]. Agricultural land boundaries were delineated using recent land use maps [17] to exclude non-cultivated areas, as recommended for precision agriculture applications [20]. The resulting SMU framework achieved a spatial resolution of 10 ha, representing a significant improvement over previous regional studies using 25 ha grids [31]. Field validation at 120 locations showed 82% agreement between mapped and observed SMU characteristics, comparable to accuracy levels reported in similar studies [32]. This terrain-integrated approach has demonstrated particular effectiveness in basaltic landscapes, with reported improvements of 20–25% in crop yield predictions [33].

This study demonstrates significant methodological advances over previous work in several key aspects. First, we achieved substantially higher spatial resolution using 10-ha sampling grids compared to the 25-ha grids employed in Bhaskar’s (2015) [17] study of the same region, enabling more precise detection of soil variability patterns. Second, our innovative integration of terrain parameters, legacy soil data, and machine learning algorithms (Random Forest with 10,000 trees) represents a marked improvement over conventional single-parameter classification approaches, as evidenced by our 28% reduction in the within-unit coefficient of variation. Third, we implemented rigorous quantitative validation protocols, including homogeneity metrics (

Table 2. Quantitative comparison of SMU delineation performance between current and previous studies in semi-arid basaltic soil.

Metric	This Study (2023)	Bhaskar (2015) [17]	Singh et al. (2017) [30]	Improvement
Spatial Resolution	10-ha grids	25-ha grids	15-ha grids	+150% vs. [17]
Within-SMU CV (%)	28.1 ± 3.2	38.4 ± 5.6	32.7 ± 4.1	26.8% reduction vs. [17]
Boundary Accuracy (κ)	0.81	0.65	0.72	+24.6% vs. [17]
Parameters Integrated	4 (Terrain + Soil + ML + Field)	2 (Soil only)	3 (Soil + Terrain)	+100% vs. [17]
Validation Points	120	80	100	+50% vs. [17]

) and field verification at 120 stratified points (κ = 0.81), which address the reproducibility limitations noted in earlier studies. These advances collectively provide a more robust framework for precision agriculture applications in semi-arid tropical regions, particularly for basaltic soil landscapes where previous methods showed limited accuracy in capturing fine-scale heterogeneity. The methodological improvements are further reflected in our ability to identify previously undetected nutrient management zones (e.g., Zn-deficient SMU13) and quantify their spatial relationships with terrain features—capabilities that were absent in prior regional studies.

2.3. Surface Soil Sampling

A stratified grid sampling schema was created using Arc-GIS version 10.3. The surface sampling was performed at a grid of 325 m × 325 m representing one sample per 10 ha area. Geo-referenced soil samples were collected using a handheld global positioning system (GPS) at a 20 cm depth from agricultural land (Figure 3). Four samples were collected and composited as one sample from each 10-ha area by using composite techniques to represent the overall fertility status of the corresponding plot. A total of 4627 soil samples were collected from the study area. As per FAO [34] guidelines, we also recorded the different landscape variables such as topography, slope, land use/land cover, parent materials, landform units, surface stoniness with drainage, and erosion status of the study area.

2.4. Soil Analysis

In the literature, characteristics such as texture, pH, organic carbon content and total nitrogen content are considered important indicators of soil quality and fertility [35]. The soil samples were air-dried and processed (<2 mm sieve) for analysis following standard laboratory methods and procedures. Soil pH and EC were determined using standard procedures [36], soil organic carbon [37], and available nitrogen (AvN) [38]. The Olsen extraction (0.5 M NaHCO₃) method for measuring available phosphorus (AvP) [39] and neutral normal ammonium acetate extraction method for available potassium (AvK) [40] were used. The sulphur (AvS) was extracted with 0.15% CaCl₂·2H₂O solution [41]. The micronutrients viz., iron (Fe), manganese (Mn), copper (Cu), and zinc (Zn) were determined using DTPA extractant [42].

2.5. Statistical Analysis

The spatial distribution of each variable was determined for a 4627-point data set. To identify outliers, a box-plots local auto-correction index of Moran [43] and its dispersion diagram [44] were used. Values which were outside the mean plus four standard deviation (SD) intervals were considered outliers and removed from further analysis. Descriptive statistics such as mean, standard deviation and coefficient of variation (CV), frequency distribution, Pearson correlation coefficient (p = 0.05), principal component, and factor analysis were carried out using EXCEL^® 2007 and SPSS 24.0^® (SPSS Inc., Chicago, IL, USA) software. Within soil mapping, a unit’s CV was used to estimate the degree of soil variability within specific land units. Soil properties with CV ≤ 15%, 15–35% and >35% were considered as low-, moderate-, and high-variability parameters [45], respectively. Threshold values of soil parameters of SMUs adopted by various researchers are given in

Table 3. Threshold values of soil fertility parameters.

Parameters	Unit	Threshold Value	Adopted by
OC	(%)	0.50–0.75	[37]
AvN	kg ha−1	280–560	[38]
AvP		22–56	[39]
AvK		140–336	[40]
AvS		10–20	[41]
DTPA-Fe	mg kg−1	≤4.5	[42]
DTPA-Mn		≤1.0
DTPA-Zn		≤0.6
DTPA-Cu		≤0.2

. In this study, geostatistical techniques were employed to analyze the spatial variability of soil fertility parameters using the geostatistical analyst extension in ArcGIS 10.3 [46]. Experimental semivariograms were computed to quantify spatial autocorrelation, and two theoretical semivariogram models, namely spherical and exponential, were fitted to the data. The selection of the best fit model was based on cross-validation statistics such as mean error (ME), root mean square error (RMSE) and the coefficient of determination (R²), following the approach suggested by Webster and Oliver (2007) [47]. The best fitted model was then used to perform ordinary Kriging interpolation, generating continuous surface thematic maps that visualize the spatial distribution of key soil properties across the study area.

3. Result

3.1. Soil Mapping Units

SMUs were delineated based on landform, soil depth, texture, and slope class, following established methodologies in pedology and digital soil mapping [48,49]. A total of 15 SMUs were identified in the study area (Figure 4). SMU1 was dominant in the alluvial plains with very gentle slopes (1–3%), very deep soils (>150 cm), and clayey texture, characteristic of depositional environments with fine-textured alluvium [50]. SMU2 to SMU5 were delineated in pediment areas with slopes ranging from very gentle to moderate (1–5%), soil depths from very shallow (<25 cm) to very deep (>150 cm), and textures ranging from clay to clay loam, consistent with lithosequence-driven soil variability [51]. SMU6 and SMU7 occurred on basaltic plateaus with shallow to moderately shallow soils and clayey textures, which are common in the Deccan Plateau region [52]. SMU8 to SMU10 were mapped in the pediplain, characterized by gentle to moderate slopes, moderately shallow to very deep soils, and clay to clay loam textures, indicative of mixed colluvial and alluvial materials [53]. SMU11 to SMU13 were associated with moderately sloping undulating terrain, where soil depths varied from very shallow to very deep, and textures ranged from clay to clay loam, reflecting topographic influence on soil development [54]. SMU14 and SMU15 were found in valley regions with moderately shallow to very deep soils and silty clay to clay textures, typical of depositional zones with higher moisture retention and finer sediments [55]. The delineation process integrated satellite image interpretation, slope analysis using digital elevation models (DEMs), and field validation, in accordance with national soil survey and mapping standards [56,57].

3.2. Variability in Soil Fertility Parameters

The SMUs varied in terms of their fertility status (Figure 5). Spatial variability maps indicate that the majority of soils were moderately alkaline and non-saline and the northern part of the study area was high in SOC content. The area is very low to low in nitrogen and phosphorus and very high in potassium content. Spatial variability maps show that more than 80% of the area was deficient in zinc and about 40% area in Fe, which needs immediate attention from soil managers for better crop production. With respect to all the SMUs, the mean pH of the soil varied from 7.6 (SMU4 & SMU7) to 8.0 (SMU11). The mean value of OC varied from 0.88% (SMU1 and SMU13) to 1.11% (SMU7 and SMU14). The mean value of the AvN was low (<280 kg ha⁻¹) in all SMUs. Moreover, AvP varied from low (15.7 kg ha⁻¹) to medium (33.1 kg ha⁻¹). AvK was very high (>336 kg ha⁻¹) in all the SMUs of the study area. The AvS varied from medium (18.3 kg ha⁻¹) to high (36.8 kg ha⁻¹). The mean values of micronutrient followed the order Mn > Fe > Cu > Zn and were sufficient in all SMUs by considering threshold limits suggested by [42] (Table 3). Variability was low for pH and AvN whereas it was high for EC, OC, AvP, AvK, AvS, and micronutrients within the SMUs. The variation in AvN was high in SMU4 (CV = 53%) (

Table 4. SMU characterization with statistical significance.

SMU	Area (%)	Depth	Texture	pH	SOC (%)	AvN (kg ha−1)	Management Implication
1	35	Very deep	Clay	7.9	0.9	121	Optimal for cotton
4	8	Deep	Clay loam	7.6	1.0	148	High N variability (CV = 53%) *
7	5	Mod. shallow	Clay	7.6	1.1	128	High SOC hotspot
13	3	Shallow	Clay loam	7.7	0.9	116	Zn-deficient (0.4 mg kg−1) **

(* p < 0.05, ** p < 0.01, ANOVA with Tukey HSD).

).

In SMU9, the variability in all the soil parameters was low except for AvK (CV = 417%). The Pearson linear correlation matrix presented in

Table 5. Pearson correlation matrix among selected soil fertility parameters.

Variables	pH	EC	OC	AvN	AvP	AvK	AvS	DTPA-Fe	DTPA-Mn	DTPA-Zn	DTPA-Cu
pH	1
EC	0.16	1
OC	−0.02	0.71 **	1
AvN	−0.49	0.12	0.47	1
AvP	−0.25	0.34	0.60 *	0.41	1
AvK	0.04	0.79 **	0.83 **	0.55 *	0.58 *	1
AvS	0.03	0.77 **	0.65 **	0.56 *	0.41	0.81 **	1
DTPA-Fe	−0.48	0.48	0.62 *	0.38	0.08	0.45	0.31	1
DTPA-Mn	−0.55	0.36	0.69 **	0.76 **	0.34	0.57 *	0.46	0.85 **	1
DTPA-Zn	0.00	0.51	0.86 **	0.26	0.68 **	0.67 **	0.37	0.45	0.52 *	1
DTPA-Cu	0.05	0.22	0.47	−0.16	−0.14	0.10	−0.15	0.66 **	0.41	0.48	1

** Significant level = 0.01, * Significant level = 0.05.

shows that OC was positively and significantly correlated with AvP (r = 0.598 *), AvK (r = 0.83 **), AvS (r = 0.65 **), Fe (r = 0.62 *), Mn (r = 0.69 **), and Zn (r = 0.86 **).

According to a variogram’s efficiency evaluation, the exponential model for an ordinary Kriging best described the semivariograms of pH, OC, and N, whereas the exponential model for a simple Kriging best defined the semivariogram of EC. The spherical model of an ordinary Kriging best described the variogram of P, K, and S. The variograms of Fe and Mn were best fitted by the exponential model for a simple Kriging, but the variograms of Zn and Cu were best defined by the exponential model for a simple Kriging. Types of kriging, best fitted model, sill, nugget, and range for soil properties are given in

Table 6. Geostatistics of the best fitted models for soil fertility parameters.

Soil Properties	Kriging Type	Fitted Model	Range (m)	Nugget (Co)	Partial Sill (C)	Sill (Co + C)	N:S ratio	RMSE **
pH	Ordinary	Exponential	1173	0.12	0.2	0.3	0.38	0.6
EC	Simple	Exponential	1599	0.15	0.17	0.3	0.47	0.1
SOC	Ordinary	Exponential	656	0.04	0.09	0.1	0.31	0.1
N	Ordinary	Exponential	608	115	308	423.0	0.27	59.0
P	Ordinary	Spherical	608	80	343	423.0	0.19	11.2
K	Ordinary	Spherical	608	120,560	176,740	297,300.0	0.41	102.8
S	Ordinary	Spherical	608	50	67	117.0	0.43	5.6
Zn	Simple	Exponential	2815	0.38	0.47	0.9	0.45	0.3
Cu	Simple	Exponential	2902	0.39	0.46	0.9	0.46	0.7
Fe	Ordinary	Exponential	1931	32	38	70.0	0.46	1.7
Mn	Ordinary	Exponential	1324	220	271	491.0	0.45	2.7

** Root Mean Square Error.

.

3.3. SMUs-Fertility Parameters Relations

To present the relationships between the studied physicochemical parameters of soils within specific SMUs, a principal component analysis was performed (Figure 6). In SMU1, two factors accounted for 51.1% of the variance. Significant correlations between AvN and Zn, SOC and AvK, Mn and Fe were observed in SMU1, while there was a poor correlation between AvK and AvS content (Figure 5). In SMU2, two factors accounted for 56.41% of the variance and high correlations were found between AvP and SOC, AvK and AvS, Mn and Cu. In SMU3, two variables accounted for 55.25% of the variance and high correlations were observed between AvN and AvP, AvK and AvS, Mn and Fe. SOC, Avk, and AvS were positively correlated to each other, but with low values. Two factors accounted for 63.25% of the variability in SMU4 and significant associations between AvK and AvP, OC and Zn, as well as AvK and AvS, were observed. Two factors explained 67, 50.53, 67.2, 59.38, 58.45, 65.59, 68.76, 54.55, 64.05, 59.04, and 54.43% of the variability in all the remaining SMUs (SMU5 to SMU15) as shown in Figure 6, respectively. Correlation between soil pH and micronutrients was negative. Within the SMUs, the variability in soil properties between the first two factors was the highest in SMU11. These data show that the area was deficient in the available nitrogen while sufficient in available phosphorus, potassium, and sulphur. The relative sufficiency levels of micronutrients across all SMUs followed a distinct descending order: Mn exhibited the highest sufficiency, followed by Fe and Cu, whereas Zn showing the lowest sufficiency among the four. This pattern indicates that, although all micronutrients were generally present at adequate levels, Mn was the most abundant, whereas Zn was comparatively the least sufficient. Notably, Zn levels remained within the sufficiency range across most SMUs, except in SMU13, where a deficiency was observed.

4. Discussion

4.1. SMU-Soil Relationships and Landscape Processes

The delineation of 15 spatially heterogeneous SMUs underscores the critical role of geomorphic processes in shaping soil variability across the study area. Our findings demonstrate a clear catena sequence (Figure 7), where erosional processes in upslope positions (SMU2-SMU5) result in shallow (<50 cm), gravel-rich soils and depositional valleys (SMU14-SMU15) accumulate deep (>150 cm), fine-textured sediments [58]. This terrain-driven differentiation was particularly evident in SMU1 (alluvial plains), where very deep (>150 cm) clayey soils exhibited optimal cotton suitability (pH 7.9, 0.88% SOC), contrasting sharply with the limitations of shallow pediment soils (SMU2) for this crop. The observed alkaline soil conditions (pH 7.6–8.0) [59], characteristic of basaltic weathering, suggest opportunities for diversifying rotations with pH-tolerant crops like sorghum (Sorghum bicolor), potentially improving system resilience compared to the current cotton−pigeon pea monoculture.

4.2. Soil Carbon Dynamics

Soils rich in OC content are an indication of a healthy soil ecosystem [60], and it correlates with different soil biodiversity aspects [61,62]. In our study, three SMUs emerged as significant carbon hotspots (SMU7, SMU14, SMU15; SOC > 1%), attributable to enhanced manure inputs (farmyard manure @ 5–7 Mg ha⁻¹ yr⁻¹) [63], residue retention practices (cotton stalk incorporation) [64], and depositional valley processes that stabilize SOC [65]. Soils in these SMUs were rich in both microbial biomass and water-holding capacity, having high plant nutrients supplying capacity and being also resistant to erosion [65,66].

4.3. Management of Nutrient Management

A comparison between critical levels and mean values of SMUs (Figure 8) provides the basic idea about the potential and constraints of each SMU, helping with a variable rate of fertilizer recommendations and judicial distribution of all soil management practices. For instance, in all SMUs the mean values of AvN were rated as low, and the AvP was rated as medium. In contrast, AvKs were rated as very high and AvSs rated medium to high, which reveals that the area was deficient in available nitrogen, while sufficient in available phosphorus, potassium, and sulphur. Deficiency in AvN is mainly due to the continuous growing of nutrient-exhaustive crops like Bt-cotton with low input management and high nitrogen loss due to volatilization in the summer season [67,68]. Thus, more application of nitrogenous fertilizer compared to other nutrient fertilizers is needed. P-fertilizer application and the formation of insoluble calcium phosphate in an alkaline soil environment led to less availability of AvP [69].

Sulphur (AvS) is built up in soils by precipitation, air, irrigation water, crop residue management, fertilizer application, etc. The losses through crop removal, co-precipitation with CaCO₃ at high pH, and erosion caused the variability in AvS status in this region [70]. Likewise, Zn was deficient in SMU13, so there is a need for more focus on the application of Zn fertilizer rather than other micronutrient fertilizers. Low variability in soil pH was observed as all soils originated from the basaltic parent material [55], with a high buffering capacity and absence of carbonate in the saturation extract [71]. The variation in AvN was high in SMU4 (CV = 53%) and might be due to factors such as management practices, soil erosion, etc., that were not considered during the delineation of SMUs. The results from [72] in south-eastern Ireland are consistent with the high CV values of AvP and AvK observed in our study area. High variability in AvP of the SMUs could be due to variation in the SOC content and the presence of residual P of long-term cultivated P fertilized fields. High SOC content in SMU14 increased the phosphorus availability. Variation in weathering status of native K-bearing minerals showed high variability for AvK in SMUs of the study area. The soil properties within SMU 9 exhibited very low variability. As it covered a very small area and samples were very homogeneous, this unit follows mono cropping with similar management. The AvK (CV = 417%) was high, and it might be due to the ubiquitous presence of potassium-bearing minerals, muscovites, and biotites [73] in association with expanding and contracting clay mineral smectites, which affect the release and retention behavior of K in the soil solution. Soil micronutrients (Zn, Mn, Cu, Fe) showed very high variability, which might be due to variation in the rainfall and soil management [74].

4.4. Precision Agriculture

The SMU-specific intervention framework (

Table 7. The SMU approach enabled targeted interventions.

SMU Group	Key Constraint	Recommended Practice	Expected Benefit
Summit (6–7)	Erosion, low SOC	Contour hedgerows + compost	40% soil loss reduction [33]
Slope (4–5)	N leaching	Polymer-coated urea	25% NUE improvement [64]
Valley (14–15)	Zn deficiency	ZnSO4 @ 25 kg ha−1	0.5–1.2 t ha−1 yield gain

) demonstrates how terrain-informed soil mapping translates to precision management. Summit positions (SMU6–7), characterized by erosion-prone slopes (>3%) and low SOC (0.8–1.1%), require structural stabilization through contour hedgerows (Leucaena leucocephala at 5 m intervals) combined with compost application (5 Mg ha⁻¹), a package shown to reduce soil loss by 40% while building organic matter [33].

For backslope units (SMU4–5) exhibiting high N leaching (CV = 53%), polymer-coated urea (PCU) applications at 80% standard N rates can improve nitrogen use efficiency by 25% by synchronizing release with monsoon patterns [64]. Valley bottoms (SMU14–15), despite their high SOC (1.1–1.3%), require targeted zinc supplementation (ZnSO₄ at 25 kg ha⁻¹ banded at sowing) to address deficiency-induced yield gaps (0.5–1.2 t ha⁻¹ in cotton), particularly where soil pH > 7.5 promotes Zn immobilization [42]. This differentiated approach could generate estimated economic gains of ₹12,000–15,000 ha⁻¹ yr⁻¹ through input savings (15–20% fertilizer reduction) and yield stabilization, based on comparable interventions in Maharashtra’s cotton systems [Sharma et al., 2022] [63].

5. Conclusions

By integrating intrinsic soil characteristics (depth and texture) and site-specific factors (slope, elevation, and landform), 15 distinct SMUs were delineated across the study area. These SMUs enhance the precision of soil management by maximizing homogeneity within units while ensuring heterogeneity between them, thereby providing a reliable framework for tailored agricultural practices. SMU1, covering 35% of the total geographical area, emerged as the dominant unit, characterized by intensive cultivation and crop diversification. Fertility assessments revealed significant variability in soil nutrients both within and between SMUs, with deficiencies in available nitrogen (AvN) and zinc (Zn) identified as key constraints to crop productivity. The integration of site-specific fertility data with spatially defined SMUs offers a robust foundation for precision farming, enabling targeted nutrient management and optimized resource use. By aligning management practices with SMU-specific conditions, farmers can reduce input costs, minimize environmental impacts, and advance toward sustainable crop production. Future studies should expand on these findings by systematically comparing SMU-driven approaches with other precision agriculture tools to refine scalable solutions.