Current legume inclusive long–term experiments were located at the ICAR-Indian Institute of Pulses Research, Kanpur (26°27′N latitude and 80°14′E longitude), India and initiated in 2003. The trial sites consisted two different cropping ecologies such included upland ecology and lowland ecology at same studied location. Upland and lowland cropping ecologies differed primarily in land topography, water drainage, soil aeration capacity, and soil properties. Upland ecologies are typically well-drained, support crops like maize and sorghum during rainy seasons (Nath et al., 2023a). On contrary, lowland ecologies are often flood-prone with poor drainage, favored water-tolerant crops such as rice (Nath et al., 2019). Soil in lowlands is anaerobic during rainy season, while upland soils are well-aerated (Raj et al., 2017). In present study, upland and lowland ecological distinctions influenced crop selection, management practices, and yield potential of crops. Based on land topography, upland cropping ecology was typically situated on higher farmland with good surface runoff and minimal water retention. In contrast, lowland ecology located in flat areas where water naturally accumulates, leading to higher moisture retention and frequent inundation, making them suitable for water-loving crops like rice. Accordingly, diversified crop rotations were designed as per ecologies in present study and K pools were studied under various cropping systems and nutrient managements. In both ecologies, four diversified crop rotations suited to the production ecology were tested (cereal-cereal and cereal-legume). Subtropical climate, with an average annual air temperature and rainfall of 26 °C and 722 mm (ranging from a minimum of 510 mm in 2015 to a peak of 1,225 mm in 2013) was climatic attributes of the study site. The soil at the trial location is classified as Fluvisol (Typic Ustochrept). Both upland and lowland ecologies exhibited sandy-loam soil texture at the 0–15 cm depth. Initial soil characteristics in the upland and lowland ecologies were presented in Supplementary Table S1.

Soil sampling was conducted in both ecologies during April 2017, coinciding with the harvest of the base crops (wheat and rice) in the 14th crop cycle with no intermediate sampling. Samples were collected from each treatment plot across three replications at two soil depths: 0–15 cm and 15–30 cm. Using a post-hole auger equipped with a sharpened edge and a 15 cm core height, samples were taekn from six randomly selected sites per plot, arranged in a Z-pattern, excluding a 1-meter margin around each plot. For each depth, the six subsamples were thoroughly mixed to form one composite sample per treatment. The collected soils were first air-dried for 72 h, sieved through a 2 mm mesh, and then oven-dried at 60 °C for 72 h prior to K fraction analysis. A total of 36 composite samples were prepared, corresponding to the experimental design (4 cropping systems × 3 nutrient management practices × 3 replications), ensuring data accuracy and consistency.

In this study, four potassium fractions namely; water-soluble K, exchangeable K, non-exchangeable K, and total K were determined. Fraction 1, i.e., Water-soluble K was determined by shaking soil with distilled water (1:10 ratio) for 1 h. Prior to centrifugation and filtration, and K-concentration was measured using a flame photometer (SYSTRONICS Model 1,027, India) (Jackson, 1973). Available K was extracted using 1 N ammonium acetate (pH 7.0) in a 1:5 soil-to-solution ratio, shaken for 5 min and assessed using flame photometer (Hanway and Heidel, 1952). Exchangeable K (Fraction 2) was deduced by subtracting water-soluble K from available K. 1 N HNO₃-extractable K was determined by boiling 1:10 soil-to-acid ratio (25 mL of 1 N HNO₃) for 10 min on a hot plate at 150–180 °C. After cooling, the mixture was filtered into a 100 mL volumetric flask and diluted to volume using 0.1 N HNO₃. The K concentration in the filtrate was measured by flame photometry, following the procedure of Wood and DeTurk (1941). Non-exchangeable K (Fraction 3) was calculated by subtracting available K from the 1 N HNO₃-extractable K. Total K (Fraction 4) was estimated by wet digestion using hydrofluoric acid (HF) and perchloric acid (HClO₄) as described by Pratt (1965). For this, 0.1 g of finely ground soil was pre-dried at 100 °C for 2 h, placed in a 30 mL platinum crucible, and moistened with a few drops of distilled water. Then, 5 mL of 48% HF and 0.5 mL of 60% HClO₄ were added. The crucible, mostly covered with a platinum lid, was heated on a sand bath at 200–225 °C until the acids evaporated. After cooling, 5 mL of distilled water and a few drops of HClO₄ were added, and the mixture was re-boiled to dryness. Once cooled again, 5 mL of 6 M HCl was added and diluted to two-thirds of the crucible volume. The solution was gently boiled until all residues dissolved and after volume make-up to 100 mL, the K content was determined using flame photometer.

Grain yield data of all component crops within each cropping system and ecology were recorded annually throughout the study. For this analysis, the correlation between soil K fractions and the yield of the base crop was assessed to understand the long-term influence of K pools on crop productivity. In upland ecology, wheat served as the base crop across all rotations, while in lowland ecology, rice was the common crop across systems. Therefore, wheat yield (upland) and rice yield (lowland) from the 14th crop cycle were used for correlation analysis. Grain yield was measured by manually harvesting a 5 m × 5 m net plot area per treatment, and the yield was standardized to 14% grain moisture content and expressed in kg ha⁻¹.

Experimental datasets were analysed following the analysis of variance (ANOVA) for split-plot design using OPSTAT program and Tukey’s Honest Significance test was used at p ≤ 0.05 for mean comparison. The correlation analysis and box plot presentation were undertaken by Microsoft Excel™ 2017. Heatmap representation was performed by online Heatmapper. Interaction effect has been assessed using t-test at 5% level of significance.

At 0–15 cm depth, maize-wheat-maize-chickpea had the highest water-soluble (15.6 kg ha⁻¹) and exchangeable K (162.7 kg ha⁻¹), while pigeonpea-wheat recorded the maximum non-exchangeable K (1826.2 kg ha⁻¹) (Table 1). Among pulse-based systems, pigeonpea–wheat showed notable improvements in non-exchangeable K by 2.3% and water-soluble K by 3.4%, although exchangeable K decreased by 6.4% compared to the maize–wheat system (p < 0.05). At 15–30 cm depth, the respective increase of non-exchangeable K and total K was 3.4, and 4.7% (p < 0.05), while exchangeable K remained unchanged under pigeonpea-wheat and maize-wheat systems. Maize-wheat-maize-chickpea systems resulted in increased level of water-soluble K (by 9%) and non-exchangeable K (by 9.3%) than maize-wheat (p < 0.05). Integrated nutrient management (INM) recorded the highest values for all K fractions than CT and RDF across depths (p < 0.05) (Table 1). Compared to RDF, INM improved water-soluble K by 37.6%, exchangeable K by 17.9%, non-exchangeable K by 12.2%, and total K by 10.1% at 0–15 cm soil depth (p < 0.05). At 15–30 cm depth, INM again outperformed, enhancing water-soluble, exchangeable, and non-exchangeable K by 25.2, 11.0, and 10.3%, respectively, over RDF (p < 0.05). Total K increased by 4.2% under INM than that of RDF (p < 0.05).

At the 0–15 cm depth, the rice–wheat system represented increase in water soluble K by 50% over rice–wheat–rice–chickpea and 30% over rice–wheat–mungbean systems (p < 0.05) (Table 2). Exchangeable K was also higher in rice–wheat by 3.4% than rice–wheat–mungbean and 13% rice–chickpea (p < 0.05). Non-exchangeable K increased under rice–wheat by 13% than rice–chickpea and 8% than rice–wheat–mungbean (p < 0.05). At 15–30 cm depth, water soluble K was higher in rice–wheat–mungbean and rice–chickpea, showing a 9% increase over rice–wheat (p < 0.05). Interestingly, exchangeable K was the highest in rice–wheat–rice–chickpea, about 10% more than rice–wheat. Non-exchangeable K followed a similar trend, with rice–wheat outperforming others, maintaining a 7% higher value than rice–wheat–mungbean and 9% more than rice–chickpea (p < 0.05). The INM significantly enhanced all K pools, with water soluble K increasing by 47% and exchangeable K by 27% over RDF at 0–15 cm soil depth (Table 2). At 15–30 cm, INM enhanced the water-soluble K (by 18%), exchangeable K (by 12%), non-exchangeable K (by 12%) over RDF (p < 0.05). Total K was significantly highest under INM across depths, improving by 8% at 0–15 cm and 7% at 15–30 cm than RDF.

Lowland ecology significantly improved K availability across soil depths compared to upland ecology (Supplementary Figure S1). At 0–15 cm, lowland soils exhibited higher concentrations of water-soluble K (~4 times), exchangeable K, non-exchangeable K, and total K. Similarly, at 15–30 cm depth, lowland conditions maintained higher levels of all K pools, especially water-soluble and exchangeable K. Specifically, water soluble K increased under lowland ecology by 249% at 0–15 cm and 230% at 15–30 cm depth (Table 3). Exchangeable K heightened by 52% at 0–15 cm but showed a non-significant change at 15–30 cm (p < 0.05). Non-exchangeable K increased by 5% at 0–15 cm and 10% at 15–30 cm depth (p < 0.05). Cropping systems combined with INM markedly enhanced K pools than with RDF. In upland ecology, at 0–15 cm depth, the maize–wheat–maize-chickpea with INM increased exchangeable K by~34% and water-soluble K by ~48–52% over maize-wheat with RDF (Figure 2) (p < 0.05). In lowland ecology, legume-based cropping systems with INM significantly improved K pools compared to rice–wheat with RDF, increasing exchangeable K by 12–18% and water-soluble K by 20–30% across depths (p < 0.05). The pigeonpea–wheat system under INM consistently recorded the higher non-exchangeable and total K in both soil layers in upland ecology than maize-wheat with RDF (Figure 3). Rice-wheat-rice-chickpea + INM, rice-chickpea + INM resulted in 5–10% increase in non-exchangeable K over rice-wheat with RDF.

After 14 years of cropping under different ecologies, soil depth significantly influenced K pools (Table 4). In upland ecology, water soluble K decreased by about 7.6% from 0–15 cm to 15–30 cm depth, while exchangeable K declined by 3.3%, both not statistically significant. However, non-exchangeable K dropped significantly by 12.7%, and total K decreased by 4.3%. In lowland ecology, water soluble K reduced by 13% with depth, exchangeable K decreased sharply by 35%. Non-exchangeable K declined by 8.2%, and total K decreased by 4.2%, both significant. Overall, K availability was higher in topsoil and lowland ecology.

In upland ecology, wheat yield showed the strongest correlation with total K (R² = 0.2417, p = 0.01), followed by exchangeable K (R² = 0.173, p = 0.02), and water-soluble K (R² = 0.1834, p = 0.02) (Figure 4). In lowland ecology, rice yield had the highest correlation with total K (R² = 0.3782, p = 0.001). This was followed by non-exchangeable K (R² = 0.1761, p = 0.02) and water-soluble K (R² = 0.1428, p = 0.03), while exchangeable K had a modest effect (R² = 0.0124, p = 0.30). Heat-map illustrated a clear interaction between cropping systems and nutrient management interaction on soil K pools (Figure 5). Across both upland and lowland ecologies, INM consistently maintained higher levels of all K-fractions, particularly under legume-inclusive systems like maize-wheat-mungbean and rice–wheat–rice–chickpea (Supplementary Figures S2, S3). In contrast, RDF treatments showed markedly reduced K levels. The 0–15 cm soil layer exhibited more pronounced K stratification than the 15–30 cm layer. Overall, INM treatments enhanced K availability, especially in lowland systems with diverse rotations.

One of the notable findings of this study is the clear advantage of lowland rice ecology in maintaining higher concentrations of all K fractions especially water-soluble K (3–4 times) and exchangeable K (by 52%) in 0–15 cm as favoured by redox state and finer soil texture cm (Sanyal et al., 2014). On the other hand, upland systems showed improved K status when legumes like pigeonpea and chickpea were included, especially under maize–wheat–maize–chickpea and pigeonpea–wheat rotations. Furthermore, ecological characteristics substantially modulate nutrient dynamics, likely due to differences in redox conditions, organic matter turnover, and microbial activity (Dey et al., 2017). However, in the upland soil more aeriation or formation of hydrogen peroxide (H₂O₂) after rainfall oxidise the structural Fe²⁺ into Fe³⁺ resulting decrement of the net negative charge in the crystal surface which ultimately weaken the soil structure and release of K (Bell et al., 2021). Similarly, other important factors like soil organic matter which has higher CEC but does not necessarily indicate higher K retention and leaching of K was significantly higher in well drained, light texture soils with heavy K application (Goulding et al., 2021; Ishfaq et al., 2023). Upland and lowland ecologies more specifically differential moisture regime have profound impact on soil K-dynamics as Zeng and Brown (2000), reported alternate wetting-drying significantly increased K⁺-fixation due to entrapment of K⁺ into deeper interlayers from the peripheral surfaces. Cereal-cereal rotations tend to mine more K, especially water-soluble and exchangeable forms, while legume-based systems can enhance exchangeable K but may rely heavily on native reserves (Singh et al., 2014; Sarkar et al., 2014). Such findings underscore the importance of site-specific (cropping system and ecology specific) nutrient management recommendations rather than a one-size-fits-all approach.

Different crops grown within a system and their respective soil fertilization methods can influence potassium (K) dynamics (Singh et al., 2014). Indian soils typically contain minerals capable of supplying substantial amounts of K to crops. However, involvement of input intensive crop cultivars insufficient K inputs can lead to significant K mining, creating large negative balances and eventually depleting the native soil K reserves (Tewatia et al., 2017). In lowland rice ecologies, higher K pools were recorded under the rice–wheat system, particularly at a depth of 0–15 cm. Singh et al. (2014) also noted that omitting K fertilization in the rice–wheat system led to the annual depletion of 158–349 kg K ha⁻¹ from soil reserves, highlighting the risk that ongoing insufficient K application could compromise long-term productivity. The superior K availability observed under the rice-wheat rotation compared to rice-wheat-mungbean or rice-wheat-rice-chickpea systems can be attributed to the differential K uptake and residue return patterns inherent to these crops. Rice and wheat residues are excellent source of K upon degradation and Li et al. (2020) underlined straw inclusion along with recommended K dose could restore non-exchangeable K but, did not guarantee rapid jump in available K. Conversely, legume-inclusive rotations might slightly reduce soil available K due to different nutrient uptake dynamics and lower residue K content. Higher K pools were observed under legume-based systems such as rice-wheat-rice-chickpea in 15–30 cm depth than rice-wheat system because of difference in root system architecture. Tap-root/deeper roots in legumes facilitates uptake of more K from the subsoil than shallow/fibrous systems of cereals crops (White, 2013). Not only that, White (2013) listed several important root mediated characters for crops with high K-acquisition efficiency and among which early root vigor and root biomass, higher root biomass and root length density, increasing affinity K-uptake capacity, affinity for K⁺-transporter protein and formation of root hairs were some of them. Therefore, cereal-legume cropping system could utilize K pools across soil profile and had higher K cycling than cereal-cereal system. Results indicated that pigeonpea inclusion, through mechanisms like deeper root systems, organic matter contribution, and enhanced microbial activity facilitating the weathering of K-bearing minerals ultimately the subsoil K pool (Nath et al., 2023a). Thus, the pigeonpea–wheat system stands out among the pulse-based rotations for enhancing non-exchangeable K especially in the subsoil layer, contributing to long-term soil fertility. However, previous literature by Høgh-Jensen and Pedersen (2003) summarized legumes were inefficient in exploring K from non-exchangeable forms than cereals due to lesser root growth, shorter root: shoot ratio and minuscule root hair formation. Nonetheless, effective planning of nutrient and crop management is essential in cropping systems to prevent nutrient imbalances in the soil. Therefore, cropping system diversification, particularly with legumes viz. pigeonpea–wheat in upland ecology and rice–wheat–rice–chickpea in lowland, showed marked effects on soil K pools in this study.

Diversified fertilizer management regimes mechanistically regulate the K in soil system (Randive et al., 2021). In present study, INM consistently outperformed the recommended dose of fertilizers for K pools across all systems and soil depths. These improvements can be attributed to the combined effects of organic amendments and balanced fertilization, which enhance soil CEC, buffer nutrient losses, and support microbial-mediated K mineralization (Sanyal et al., 2014). The beneficial impacts of residue incorporation on exchangeable K was due to (i) residue act as insulator preventing soil heat flux, (ii) hinder K leaching through mass-flow, (iii) augment microbial activity via supplying labile C sources, (iv) the decomposed organic residue preferably attached with high valent cations (Al³⁺, Mg²⁺ and Ca²⁺) leaving behind K⁺ ions in the exchangeable surfaces and further for the crops (Sharma et al., 2023). Residue quality is also an important consideration to ascertain nutrient release kinetics and biochemical characteristics like total C and N in the residue, C: N ratio, lignin: N ratio and (lignin+polyphenol)/N were useful criterion to characterize residue degradation (Partey et al., 2013). Partey et al. (2013) reported among the cations, K has the fastest release rates and K-release from plant residue do not necessarily correlated with residue decomposition as it is not present as an organic form in plant. Potassium release study from three different residue combination showed, maximum K release occurred within 2 weeks of incubation and there was a significant difference between sole maize, Mexican sunflower and faba bean with maximum K-release under mixed residue (maize+ faba bean) (Partey et al., 2013). Laboratory incubation study by Jalali (2011) corroborates similar findings as results highlights released K had positive correlation with residue K content (r = 0.998) and negatively correlated with total C: K ratio of the residue (r = 0.990). While, microorganisms contribute to K availability by organic acid mediated solubilization of soil minerals (mica, illite) or chelating the Si-ion to make a way for the K into solution (Basak et al., 2021; Basak et al., 2017). The clear superiority of INM indicates its potential as a sustainable alternative to sole chemical fertilizer application, which can deplete soil reserves over time. While chemical inputs may improve short-term availability, they often degrade soil health and K balance. Conversely, INM enhances long-term K sustainability by combining chemical, organic, and biofertilizer sources (Hazra et al., 2019). Therefore, K application based on soil available K might misguide external K application recommendation because of the response of plants to a varied dynamic nature of K pools over time (Sanyal et al., 2014).

Soil depth also played a key role in K stratification. Both ecologies showed a general decline in K availability from 0–15 cm to 15–30 cm depths, with significant reductions in non-exchangeable and total K, particularly in lowland soils. This suggests surface accumulation of nutrients due to limited vertical movement and highlights the need for shallow-rooted crops to capitalize on surface nutrient enrichment (Das et al., 2018). Furthermore, application of fertilisers and crop residues periodically in the soil surface prop up the levels of soil K pools unlike the sub-surface layer (15–30 cm) (Islam et al., 2023). Lastly, the positive correlations between K fractions and crop yields-especially total K with rice (R² = 0.3782) and wheat (R² = 0.2417), confirm that reserve K pools contribute to yield stability in long-term systems. These results emphasize the importance of maintaining adequate non-exchangeable and total K to sustain productivity.

While the study provides valuable long-term insights, it is constrained by its site-specific nature and limited cropping systems. Broader validation across diverse soil types, climate regimes, and management intensities is needed. Moreover, mineral K reserves (structural K) were not assessed, which could provide a more comprehensive view of total K sustainability. Future strategies should focus on K budgeting, precision K application technologies and enhanced organic inputs (e.g., engineered biochar, compost blends and sewage sludge) to further stabilize the K reserve pool. Also in this particular study, assessment of mineralogical changes under long term cropping practices using X-ray diffraction analysis (XRD) was missing and should be included in the future efforts. Moreover, economics analysis is needed to check the profitability of legume inclusion under long run system.

The scalability of these findings is high, particularly in agro-ecological zones of South Asia, where; resource-poor farmers often face K deficiency due to nutrient mining. The study provides a robust framework for integrating crop and nutrient strategies into location-specific fertilizer recommendations. Specific policy and recommendations which would benefit the farming community are: i) Irrespective of production ecologies it is critical to introduce at least one pulse crop in intensive cereal based system to improve K availability specifically in the IGP ii) in long-run cropping systems like Rice–wheat–rice–chickpea or pigeonpea–wheat, combined with INM, significantly sustains non-exchangeable K pools, prevents mineralogical changes and overall soil quality.

In summary, this study provides strong scientific support for adopting INM and legume-based crop rotations tailored to ecological conditions as a strategy to maintain K sustainability vis-à-vis succour prolonged agricultural productivity. INM combined with legume-based diversified cropping systems significantly enhances K availability across soil depths and ecologies. Overall, this study provides a scientific basis for developing regionally optimized, ecologically aligned, and sustainable potassium management strategies in intensive cropping systems (Supplementary Figure S4).