The test variety was F You 498, a hybrid indica rice cultivar with a total growth duration of 155 days. The preceding crop was wheat. Field experiments were conducted during the 2018 and 2019 growing seasons at the Modern Agricultural Science and Technology Park of Sichuan Agricultural University, located in Chongzhou, Chengdu, Sichuan Province (30°42′ N, 103°28′ E). The surface soil layer (0–20 cm) was classified as sandy loam, and baseline soil fertility parameters are shown in Table 1 .

Meteorological data for the rice growth periods in 2018 and 2019 were obtained from the Sichuan Meteorological Bureau ( Figure 1 ).

A split-plot design with two factors was implemented: main plots were assigned to two K application regimes (basal:panicle = 10:0 [K1] and 5:5 [K2]), and subplots to three N splitting regimes (basal:tillering:panicle = 7:3:0 [N1], 5:3:2 [N2], and 3:3:4 [N3]). Fertilizers included urea (N≥46%), calcium superphosphate (P₂O₅≥12%), and potassium chloride (K₂O≥60%). Total N and K inputs were uniformly 150 kg ha^-1. Basal N and K were applied pre-transplanting, tillering fertilizer at the 5-leaf stage, and panicle fertilizer split equally at the 4th and 2nd leaf tip emergence stages. Phosphorus (P₂O_5–75 kg ha^-1) was broadcast entirely as basal fertilizer.

Pot-mat seedlings were machine-transplanted on May 16, 2018, and May 20, 2019, at a hill spacing of 30 cm × 18 cm with 2–3 seedlings per hill. The experiment comprised 18 plots (three replicates), each covering 15 m². A 30-cm-wide and 20-cm-high soil ridge wrapped in plastic film separated the plots to enable independent water and fertilizer management. All other practices followed conventional field management protocols.

Microsoft Excel 2016, SPSS 25.0 (SPSS Institute Inc., Chicago, USA), and Origin Pro 2020 (OriginLab, Northampton, MA, USA) were used for data analysis and visualization. Differences among treatments were tested using the least significant difference method at P < 0.05.

Year, K application, and N application rate significantly or highly significantly affected the number of effective panicles, grains per panicle, and grain yield ( Table 2 ). The year × K interaction significantly influenced the number of effective panicles, while the year × N interaction significantly affected grains per panicle. The K × N interaction highly significantly impacted the number of effective panicles and grain yield.

Across K treatments, grain yield was higher in K2 than in K1, with an average increase of 3.06% over two years. Under K2, yield increased progressively with panicle N application (N3 > N2 > N1), showing significant differences: N3 increased yield by 12.17% compared with N1 and by 4.77% compared with N2.

Regarding yield components, the number of effective panicles, grains per panicle, total spikelets, and grain-filling percentage were generally higher in K2 than in K1, though most differences were not statistically significant. Under K2, the number of effective panicles, grains per panicle, and total spikelets peaked in N3, significantly exceeding N1 by 7.57%, 11.18%, and 19.57%, and N2 by 3.26%, 3.89%, and 7.37%, respectively. Grain-filling percentage showed a declining trend, with relatively significant differences among treatments.

Different small letters after the data in the same column indicate significant differences between treatments under the same potash treatment and the same year, and the capital letters after the average data indicate significant differences between K1 and K2 (P<0.05) * and ** indicate significant effects at the 0.05 and 0.01 probability levels, respectively, and ns indicates no significant effect.

Year and K application highly significantly affected post-heading dry matter translocation ( Table 3 ). N application rate and the K × N interaction significantly or highly significantly influenced total dry matter accumulation and post-heading translocation. The three-way interaction (year × K × N) significantly affected the contribution rate of post-heading translocation.

Across K treatments, total dry matter accumulation at various growth stages mostly showed nonsignificant differences. Post-heading translocation was higher in K2 than in K1, with mostly significant differences among treatments. On average, K2 increased translocation proportion and contribution rate by 5.29% and 7.64%, respectively, compared with K1 over two years. Under K2, total dry matter accumulation at all growth stages peaked in N3, with significant differences observed. At maturity, N3 increased accumulation by 19.02% compared with N1 and by 9.59% compared with N2. Post-heading translocation also peaked in N3, exceeding N1 in translocation amount, proportion, and contribution rate by 25.54%, 5.37%, and 7.42%, and N2 by 12.68%, 2.76%, and 2.57%, respectively.

Year, N application rate, and the interactions of year with K or N application significantly or highly significantly affected stem–sheath dry matter output rate, stem–sheath dry matter conversion rate, leaf dry matter output rate, and leaf dry matter conversion rate ( Table 4 ). K application highly significantly influenced stem–sheath dry matter output rate and stem–sheath dry matter conversion rate. The three-way interaction (year × K × N) highly significantly affected stem–sheath dry matter output rate and stem–sheath dry matter conversion rate, and significantly affected leaf dry matter output rate.

Across K treatments, stem–sheath dry matter output rate, stem–sheath dry matter conversion rate, and harvest index were higher in K1 than in K2. On average, K1 increased these indices by 19.69%, 22.40%, and 1.61%, respectively, compared with K2 across two years. Under K2, stem–sheath dry matter output rate and conversion rate were highest in N1 (N1 > N2 > N3). Conversely, leaf dry matter output rate, leaf dry matter conversion rate, and harvest index peaked in N3, with mostly significant differences. N3 increased leaf dry matter output rate, leaf dry matter conversion rate, and harvest index by 45.34%, 53.64%, and 4.52% compared with N1, and by 22.12%, 22.30%, and 1.86% compared with N2.

Year highly significantly affected N accumulation at the jointing and heading stages ( Table 5 ). K application significantly influenced N accumulation at peak tillering, heading, and maturity. N application rate highly significantly affected all indices except N accumulation at peak tillering. The K × N interaction significantly or highly significantly impacted N accumulation at all stages and the N harvest index. The year × N interaction and the three-way interaction (year × K × N) highly significantly affected N accumulation at jointing.

Across K treatments, plant N accumulation during growth stages was higher in K2 than in K1, although differences were nonsignificant. At maturity, K2 averaged a 4.11% increase compared with K1. N productivity (dry matter basis) was higher in K1, while the N harvest index was higher in K2, though both differences were nonsignificant. Under K2, N accumulation increased with N application (N3 > N2 > N1), showing significant differences: at heading and maturity, N3 increased accumulation by 36.59% and 31.27% compared with N1, and by 16.16% and 14.69% compared with N2. Grain N productivity decreased with increasing N application (N1 > N2 > N3), with significant differences. The N harvest index peaked in N3, significantly exceeding N1 by 1.70% and N2 by 1.01%.

Year highly significantly affected all N export parameters in machine-transplanted rice ( Table 6 ). K application, N application rate, and their interaction highly significantly influenced N export amount and increased grain N. The year × K and year × N interactions significantly or highly significantly affected N export amount and the export contribution rate. The three-way interaction (year × K × N) significantly or highly significantly impacted export contribution rate and increased grain N.

Across K treatments, most N export parameters were higher in K2 than in K1. On average, N export amount and increased grain N rose by 8.44% and 5.01%, respectively, compared with K1 over two years. Under K2, N export amount, export rate, and increased grain N increased with higher panicle N application, peaking in N3 (N3 > N2 > N1). N3 increased these parameters by 38.09%, 4.62%, and 27.45% compared with N1, and by 14.53%, 1.45%, and 12.45% compared with N2. The N export contribution rate showed inconsistent trends over the years.

Year highly significantly affected total K accumulation, K dry matter production efficiency, and K harvest index at late growth stages ( Table 7 ). K application highly significantly influenced total K accumulation at maturity and K harvest index. N application significantly or highly significantly impacted total K accumulation at all stages and K dry matter production efficiency. The year × K and year × N interactions significantly or highly significantly affected the K harvest index. The K × N interaction highly significantly influenced total K accumulation at all stages. The three-way interaction (year × K × N) highly significantly affected K grain production efficiency.

Across K treatments, total K accumulation during late growth stages was higher in K2 than in K1, averaging a 3.46% increase at maturity. K dry matter production efficiency was highest in K1, though differences among treatments were nonsignificant. The K harvest index peaked in K2, averaging an 8.49% increase compared with K1. Under K2, total K accumulation peaked in N3, differing significantly from N1 and N2: at heading and maturity, N3 increased accumulation by 21.82% and 26.94% compared with N1, and by 8.83% and 11.90% compared with N2. K dry matter production efficiency peaked in N1 with significant differences, while the K harvest index showed inconsistent trends between years.

Year and K application highly significantly affected K export amount, export rate, and increased grain K in machine-transplanted rice ( Table 8 ). N application rate, the year × K interaction, the year × N interaction, and the K × N interaction significantly or highly significantly influenced all K export parameters.

Across K treatments, K export amount and increased grain K were higher in K2 than in K1, averaging increases of 6.85% and 13.70%, respectively, compared with K1 across two years. Under K2, K export amount and increased grain K increased with higher panicle N application, peaking in N3. N3 increased these parameters by 11.46% and 28.26% compared with N1, and by 13.35% and 18.35% compared with N2. K export rate and contribution rate showed inconsistent trends between years.

Year significantly or highly significantly affected internode length, wall thickness, and plant height in machine-transplanted rice ( Table 9 ). K application highly significantly influenced the wall thickness of S2 and S3 internodes. N application and the year × K interaction significantly or highly significantly affected wall thickness. The K × N interaction significantly or highly significantly impacted internode length and wall thickness. The three-way interaction (year × K × N) highly significantly affected S4 internode length and the wall thickness of S2 and S4 internodes.

Across K treatments, S2 internode length and S4 stem diameter were higher in K2 than in K1, although differences were nonsignificant; plant height was higher in K1 than in K2, also with nonsignificant differences. Under K2, S2 and S4 internode lengths peaked in N3, while S3 internode length peaked in N1. S2 stem diameter peaked in N3. S2 wall thickness peaked in N3, while S3 and S4 wall thickness peaked in N2, all with significant differences.

Year and K application highly significantly affected internode resistance and breaking moment of all internodes ( Table 10 ). N application significantly or highly significantly influenced all parameters except the S4 bending moment. The year × K interaction significantly or highly significantly impacted the S2 and S3 bending moments and the breaking moment. The K × N interaction significantly or highly significantly affected the S2 and S3 bending moments and the S4 breaking moment.

Across K treatments, internode resistance and breaking moment were higher in K1 than in K2. The lodging index of S2 and S3 was higher in K2 than in K1, while that of S4 was higher in K1 than in K2; all differences were nonsignificant. Under K2, internode resistance followed the trend N3 > N1 > N2, with significant differences: N3 increased resistance by 3.53% (S2), 3.16% (S3), and 4.01% (S4) compared with N1, and by 7.76%, 8.38%, and 8.92% compared with N2. The breaking moment of all internodes peaked in N3, with significant differences. The lodging index showed the trend N1 > N2 > N3, with significant differences.

Yield showed significant positive correlations with total dry matter accumulation, N accumulation, the N harvest index (NHI), and K accumulation, but significant negative correlations with rice grain N productivity and rice grain K productivity. Both dry matter N productivity and rice grain N productivity, as well as dry matter K productivity and rice grain K productivity, were significantly positively correlated with the S4 lodging index. Additionally, the S2 lodging index was significantly positively correlated with the S3 lodging index.

Nitrogen and potassium, as essential macronutrients for rice growth, critically influence yield and its components. Our results demonstrate that under one-time basal potassium application (K1), moderate panicle nitrogen (N2) significantly increased the number of productive panicles and total spikelets by coordinating nutrient allocation between tillering and panicle initiation stages. This effect is attributed to optimized dry matter accumulation from peak tillering to heading under N2, which enhanced photosynthate translocation to panicles (Li et al., 2025). However, split potassium application (K2) altered nitrogen response mechanisms. Under post-anthesis potassium postponement, high panicle nitrogen application (N3) synergistically enhanced yield through three integrated physiological mechanisms: (i) delayed potassium application precisely matched the peak potassium demand during panicle differentiation, activating nitrate reductase and pyruvate kinase activity to promote photoassimilate translocation to panicles, resulting in a 25.54% increase in post-heading dry matter remobilization (Shova et al., 2023); (ii) increased panicle N application delayed root senescence during grain filling and improved N assimilation efficiency, elevating spikelets per panicle by 11.18%; and (iii) split potassium application regulated cell turgor pressure to optimize vascular bundle development, effectively mitigating canopy overcrowding induced by high nitrogen inputs (Islam and Muttaleb, 2016). Insufficient spikelet number constitutes a primary yield-limiting factor in machine-transplanted rice. Potassium supplementation effectively optimizes nutrient supply and physiological conditions during this critical phase, promoting spikelet differentiation while reducing degeneration, thereby significantly increasing grains per panicle (Zhang et al., 2010). Concurrently, enhanced panicle nitrogen synergized with delayed potassium application to boost dry matter accumulation and translocation efficiency during grain filling, ultimately enlarging sink capacity and strengthening source flow for substantial yield improvement.

Regarding yield components under different N–K management strategies, the K2N3 regime exhibited superior performance in productive panicles, spikelets per panicle, and total spikelets. This superiority likely stemmed from improved root uptake of water and nutrients by panicle-stage N–K application, which enhanced tillering capacity and plant vigor. Furthermore, nitrogen application prolonged the active grain-filling phase (Jiang et al., 2022), while potassium regulated filling rate to synergistically increase grain weight (Liu et al., 2024), explaining the yield advantage of K2N3 observed herein. Nevertheless, this combination reduced the seed-setting rate, potentially due to excessive vegetative growth from increased nitrogen proportion at the booting stage, which disrupted nutrient allocation during anthesis and impaired pollination. This observation corroborates the conclusion that N–K imbalance may restrict grain-filling completion through sink–source limitations (Chen et al., 2024). Simultaneously, our analysis demonstrates that optimized N–K stoichiometry significantly enhances economic returns through yield improvement, particularly in the K2N3 treatment, which could effectively increase farmers’ income in local agricultural systems.

Nitrogen uptake and translocation efficiency constitute the core physiological basis for rice yield formation. Studies indicate that potassium activates nitrate reductase activity, enhancing nitrogen assimilation and partitioning to panicles (Gao et al., 2019) while reducing nitrogen retention in stems and sheaths. Topdressing 50 kg ha^-1 K at the jointing stage significantly increased plant nitrogen accumulation at flowering, grain nitrogen accumulation at maturity, and the nitrogen translocation contribution rate (Guo et al., 2016). The split application of potassium fertilizer further enhanced nitrogen translocation efficiency by 38.09% through activation of H⁺-ATPase pumps that drive nitrate transporters. The underlying physiological mechanism involves potassium serving as an activator of nitrate reductase, thereby promoting the allocation of nitrogen assimilation products to panicles while reducing nitrogen retention in stems and leaf sheaths (Vijayakumar et al., 2022). When 50 kg ha^-¹ of potassium was top-dressed at the jointing stage, the treatment significantly increased plant nitrogen accumulation at flowering, grain nitrogen accumulation at maturity, and nitrogen use efficiency (Wang et al., 2025). Concurrently, machine-transplanted rice exhibits significantly elevated nitrogen uptake from heading to maturity. Appropriately increasing the proportion of panicle nitrogen promotes root vitality, facilitating more efficient soil nitrogen absorption and accumulation (Hu et al., 2018; Zheng et al., 2023). Panicle nitrogen application precisely matches the critical period from panicle initiation to grain filling, enhancing shoot nitrogen accumulation and translocation capacity from stems and sheaths to leaves, while increasing nitrogen recovery efficiency by 7.27%–26.06% (Zhang et al., 2022). Thus, split potassium application potentiates nitrogen metabolism, fully unlocking the yield potential of panicle nitrogen to synergistically achieve efficient nitrogen supply for grain sink filling.

The panicle initiation to heading phase represents the peak potassium uptake window, where potassium supply efficacy directly governs subsequent K translocation to grains (Yu et al., 2012). Under the K2N3 regime, potassium accumulation in mature plants increased by 18%–25%. The molecular mechanism involves panicle nitrogen application activating the expression of potassium channel proteins and transporters, which enhances root potassium uptake capacity while optimizing vascular bundle potassium translocation efficiency (Vijayakumar et al., 2024). This finding is corroborated by studies in Iranian rice systems, which demonstrated that in water-scarce regions, application of 120 kg ha^-1 nitrogen combined with 80 kg ha^-1 potassium comprehensively improved rice osmoregulation capacity, protected photosynthetic organs, and ultimately enhanced drought resistance while maintaining grain yield (Rabiei et al., 2021). However, excessive potassium application can induce potassium retention in vegetative tissues, inhibiting redistribution to grains. Optimal K application rates have been reported at 108–120 kg ha^-1 (Hou et al., 2018). Notably, panicle nitrogen systematically enhances fertilizer use efficiency and significantly reduces agricultural greenhouse gas emissions (Poudel et al., 2025) by activating root uptake capacity, strengthening vascular bundle transport efficiency, and optimizing inter-organ allocation strategies. Therefore, by synchronizing nutrient supply with uptake peaks and improving redistribution, K2 synergistically interacts with N3 to systematically enhance productivity, thereby establishing the physiological foundation for efficient potassium accumulation in grains and yield improvement.

Machine-transplanted rice exhibits significantly weaker lodging resistance than manually transplanted rice due to inferior stem solidity and root anchorage, necessitating precise N–K co-regulation to synergistically enhance morphological and mechanical traits (Sun et al., 2017). This study is the first to elucidate the mechanism at the level of cell wall construction: under split potassium application (K2), the wall thickness and bending resistance of the S2 internode significantly increased. This improvement is attributed to potassium ions activating the phenylpropanoid pathway, which promotes lignin monomer synthesis and cell wall cross-linking (Ding et al., 2021). Under the K2N3 treatment, the wall thickness and stem diameter of both S2 and S4 internodes increased while shorter basal internodes, directly improving stem bending resistance. At the mechanical level, this combination significantly enhanced bending moment resistance by increasing cellulose deposition and vascular bundle development, thereby minimizing the lodging index. This finding is consistent with Zhang et al. (2021), who reported that “increased potassium application significantly reduces the lodging index by increasing lignin content or cell wall thickness in rice stems.” Moreover, split nitrogen application optimized the enrichment of photosynthetic products and nitrogen in the basal stems, supporting early-stage population establishment while avoiding stem weakness caused by late-stage nitrogen excess, thereby preventing yield losses due to lodging stress (Lu et al., 2025). However, excessive basal nitrogen application (e.g., K1N1) induced gibberellin synthesis, which promoted internode elongation and diluted cell wall material concentration, ultimately reducing mechanical strength.

As essential macronutrients for plants, adequate nitrogen and potassium directly promote the accumulation of photosynthetic products and assimilate translocation, laying the material foundation for high yields. However, under high-yield conditions, grain output efficiency per unit nutrient decreases, which may stem from the “yield–efficiency trade-off” effect, whereby nutrient use efficiency tends to decline as yields approach higher levels. Regarding stem lodging resistance characteristics, the lodging indices of S2 and S3 internodes showed a significant positive correlation ( Figure 2 ), possibly because vascular bundle development in basal internodes is continuous. When the cell wall thickness of the S2 internode decreases, the resulting reduction in bending resistance increases the mechanical load on upper internodes (Yuan et al., 2025). Notably, the dry matter productivity of nitrogen and potassium showed a significant positive correlation with the lodging index of the S4 internode. This result is the first to reveal the intrinsic relationship between nutrient use efficiency and stem mechanical properties. We speculate that the mechanism may involve plants preferentially allocating more resources to grains rather than to the synthesis of stem structural components (e.g., cellulose and lignin) when nitrogen and potassium use efficiency is excessively high, thereby reducing stem mechanical strength. These findings provide new evidence supporting the “yield–lodging resistance synergy” theory (Zhang et al., 2025).

Although this study systematically analyzed the regulatory mechanisms of nitrogen–potassium coordination on yield formation and stress resistance in machine-transplanted rice, some limitations remain. First, the conclusions are based on short-term observations of specific indica rice varieties, and their applicability to japonica ecotypes and long-term cultivation requires further validation. Second, although the improvement in stem lodging resistance is associated with potassium management, key evidence on cell wall component reconstruction (e.g., spatial deposition patterns of lignin) remains insufficient. Third, the environmental risks under high panicle fertilizer application were not assessed. Future research should combine long-term, multi-ecological field trials to deepen the synergistic analysis of post-anthesis nutrient translocation barriers and stem microstructure development, and establish an environmental footprint accounting system to improve the sustainable application of nitrogen–potassium management in machine-transplanted rice.