The long-term field investigation on conservation agriculture (CA)-based maize–mustard rotation (MMR) was initiated in the rainy season of 2013 at the ICAR-Indian Agricultural Research Institute, which is situated at 28°38′ N; 77°11′ E, New Delhi, India. The experimental location has a sub-tropical, semi-arid environment, distinguished by its scorching summers and chilly winters. The average annual precipitation fluctuates between 697 and 1,541 mm, of which ~80% occurs between July and September. The minimum and maximum temperatures and relative humidity ranges between 5°C–26°C, 20°C-40°C, and 65–83%, respectively (Supplementary Table 1). Maize was grown from July to October, and mustard is grown from October to April every year. According to pre-experiment chemical analysis of the upper (0.00–0.15 m) soil depth, the soil was alluvium-derived sandy loam (Typic Haplustepts) with a pH of 7.3, Walkley-Black C (oxidizable SOC) 0.40%; KMnO₄ oxidizable N 159.9 kg ha⁻¹ Subbiah and Asija (1956); 0.5 M NaHCO₃ extractable P 15.6 kg ha⁻¹ Olsen et al. (1954a,b), and 1 N NH₄OAc extractable K 161.3 kg ha⁻¹ Hanway and Heidel (1952).

The combinations of different tillage and nutrient management practices were assessed for the 9 years (2013–21) continuously in a fixed plot employing the split-plot design and replicated thrice. Each experimental unit was 20.0 m × 8.5 m in size, resulting in a 170 m² area. Tillage practices (03), i.e., double zero-tilled flatbed (ZTFB), permanent beds (PNB), and conventional tillage (CT), were allocated in the main plots, and nutrient management practices (03), i.e., (i) farmers’ fertilizer practice followed by, the recommended fertilization (RDF): FFP (7y) fb RDF (2y)—(⅓ N basal + surface banding of ⅓ N at knee-high stage and ⅓ N at tasseling); (ii) RDF followed by RDF improved: RDF (7y) fb RDF_I (2y)—[⅓ N basal + ⅓ N sub-surface banding (SSB) at knee-high and ⅓ N surface banding at tasseling]; (iii) nutrient expert-guided (NE) fertilization followed by NE improved: NE (7y) fb NE_I (2y)—(⅓ N basal + ⅓ N SSB at knee-high and ⅓ N surface banding at tasseling) were allocated in the sub-plots (Table 1). After harvesting of the crops in each season, all the aboveground residues were removed from the conventional plots, whereas, in the ZT and PNB plots, maize stubbles and mustard stalks (~2.5 Mg ha⁻¹ on a dry weight basis, each crop) were recycled continuously for the 9 years (Figures 1a, 1b; Supplementary Figure 1); the excess stover/stalks were utilized for animal feeding and household purposes.

In the CT plots, the experimental field was deep-ploughed to 0.3 m using a disc plough and then pulverized twice with a harrow, followed by leveling using a laser-equipped land leveler for the final seedbed preparation. Similarly, in the PNB (0.30 m furrow and 0.37 m bed width), reshaping and sowing were performed simultaneously using a raised bed planter, followed by packing using a disc coulter. In contrast, in the ZTFB, the crop seeds were drilled using a drill equipped with inverted T-type furrow openers (Supplementary Figure 2). During each season, the high-quality maize genotype ‘HQPM 1’ was sown in the 1^st fortnight of July at a spacing of 0.67 m × 0.20 m using 20 kg seed ha⁻¹, but later this genotype was replaced by the recent ‘PMH-1’ and ‘PJHM-1’ genotypes during the years 2017 and 2021, respectively. In contrast, after harvesting maize in early October, the popular mustard genotype ‘Pusa Vijay’ (4 kg seed ha⁻¹) was sown at the second fortnight of October every year. A description of the fertilizer management protocols or practices used in each treatment is described in Table 1. At sowing, the maize crop received ⅓ of nitrogen (N), total phosphorus (P), and potassium (K) as basal in the form of urea CO (NH₂)₂, di-ammonium phosphate ((NH₄)₂HPO₄), and potassium chloride (KCL), respectively. The remaining N was applied as per treatment through subsurface banding at knee-high (KH) and tasseling (TS). However, in mustard, half N was applied at the seeding, and the half was top-dressed at the first watering. Since a nutrient expert (NE) for the mustard is not available, the fertilizers were given as per the RDF in NE_I-based units. In conservation tillage (ZT or PNB), glyphosate (1 kg a.i. ha⁻¹) was sprayed 1 week before seeding, followed by the application of atrazine/pendimethalin (as pre-emergence, 1–3 days after sowing). In contrast, in the CT plots, weeds were controlled by two- to three-hand weeding. Irrigation water was applied considering the quantity of precipitation and duration of the dry spell during the cropping cycles. Furthermore, the total quantity of water (precipitation + irrigation) applied in maize–mustard rotation was 1,561, 1,167, 1,032, 1,483, 1,045, 1,042, 1,081, 1,117, and 1758 mm in the respective years during the experimentation. In general, mustard is irrigated two to three times per season, whereas maize is irrigated four to five times per season. Need-based plant protection measures were employed to control insect pests and diseases.

The samples for yield assessment were taken from a central net plot (9 × 8 m = 72 m²) leaving border rows from each side of the experimental plot. During every October and April, the maize and mustard crops were harvested, leaving ~0.20 m and ~ 0.45 m stubbles from the ground level, respectively. After harvesting, the produce was sun-dried, threshed, and cleaned, and the moisture content was adjusted to ~15% for maize and ~ 12% for mustard to report the grain yield (Ghosh et al., 2022). Furthermore, the stover/stalk yields were measured by deducting the grain weight from the respective total biomass yield (Mg ha⁻¹). Additionally, for a detailed comprehension of yield efficiency on a cropping system level, individual crop yields were converted to system productivity in terms of maize grain equivalents (EY) at the market price scale using the formula described by Biswakarma et al. (2021).

Wherein, EY = Grain equivalents of maize (Mg ha⁻¹), Y_maize = Grain yield of maize (Mg ha⁻¹), Y_mustard = Seed yield of mustard (Mg ha⁻¹), P_maize = Maize grain price (US$ Mg⁻¹), and P_mustard = Mustard grain price (US$ Mg⁻¹).

The Kjeldahl method (DISTYL EMS, KEL PLUS) using alkaline permanganate was used to determine soil nitrogen (Subbiah and Asija, 1956). Nitrate-N (NO₃⁻-N) and ammonium-N (NH₄⁺-N) were quantified via UV spectrophotometry (U128, Hitachi, Fukuoka, Japan). Briefly, 10 g of soil was combined with 100 mL of 2 M KCl and shaken on a mechanical shaker for 1 h at 200 r/m and 25°C, and then the supernatant was filtered. The NO₃⁻ -N and NH₄⁺-N contents were determined using a suitable volume of aliquot of the above extract following the procedure described by Jackson (1958) and Schuffelen et al. (1961), respectively.

A serial dilution method and standard plate count method were used to count the different groups of microbes. A nutrient agar medium was used for total bacterial counts (TBCs), potato dextrose agar medium for total fungal counts (TFCs), and actinomycete isolation agar medium (HiMedia, India) for actinomycetes. A fumigation extraction method was used to measure soil microbial biomass-C (MBC) (Vance et al., 1987). Pre-weighed soil samples from various soil depths were fumigated for 24 h with ethanol-free chloroform. The non-fumigated set was also maintained separately. Furthermore, a solution of 0.5 M K₂SO₄ (soil to extractant ratio of 1: 4) was added and subjected to reciprocal shaking for a duration of 30 min and subsequently filtered using a filter paper, i.e., Whatman No. 42. The soil OC content of the filtrate was determined using the dichromate digestion method followed by back titration with a (p = 0.05) N solution of ferrous ammonium sulfate. The MBC was then determined using the provided equation:

Where, EC = (Corg in fumigated soil–Corg in non-fumigated soil), and expressed in μg C g⁻¹ soil.

Soil microbial biomass-P (MBP) was measured using the procedure (Brookes et al., 1982). The fumigation extraction process was performed as in MBC. Each soil sample was duplicated; one was fumigated, and the other was not. Olsen’s method (Olsen et al., 1954a,b) was then used to estimate soil-available phosphorus from both fumigated and non-fumigated samples. The soil MBP was determined by subtracting the phosphorous values of the fumigated and non-fumigated samples.

The water velocity was monitored using a water meter placed in the main channel. In general, the maize and mustard crops received 3–5 and 2–3 irrigations, respectively, based on the critical growth stages and occurrence of the rainfall events during the cropping cycle. The amount of total water and depth was computed by equations as suggested by Jat et al. (2009).

Where, F is the flow rate (m³ s⁻¹), t is time (s) taken in each irrigation in each plot, and A is plot area (m²).

Additionally, the data pertaining to the amount of rainfall received during the growing season was obtained from the neighboring agro-meteorological station (Division of Agricultural Physics, IARI, New Delhi). The effective rainfall was calculated using the established methodology outlined by Mohammad et al. (2018). The cumulative irrigation input for each experimental unit was determined by adding the amount of irrigation water applied and the effective rainfall. The water productivity, expressed in kg grains ha⁻¹ mm⁻¹ of water, was calculated using the method outlined by Bhushan et al. (2007).

Water productivity = grain yield kg ha − 1 / total water applied mm

Furthermore, the system’s water productivity was determined by adding the water productivity of the maize and mustard crops. The Parihar et al. (2022) formula was used to calculate the water footprint (WF).

Energy is the key to optimizing the crop productivity and farm income. In crop production, energy input includes both direct (i.e., labor, fuel, and electricity) and indirect (i.e., seed, fertilizers, and chemicals) sources. This was later categorized into the renewable and non-renewable energy sources for further computation and interpretation. The energy balance of the maize–mustard rotation was determined using the standard conversion equivalent given by Saad et al. (2016) and the energy output by multiplying the corresponding energy coefficients with the system yield (Jat et al., 2020). To evaluate the energy efficiency of individual management practices, various energy indices were calculated using the (Equations 1–10) (Ghosh et al., 2022).

An analysis of variance was used to determine treatment effects (Gomez and Gomez, 1984). A post hoc test of the difference of means was performed using Tukey’s honestly significant difference test (p = 0.05) using SAS 9.3 (SAS Institute, Cary, NC). A pooled analysis was performed because the coefficients of variance of the main and interaction effects varied significantly between years, with no significant year and treatment interactions.

The years 2021–22 had the most precipitation, 1,608 mm from July to April, followed by 1,368 mm during 2013–14 and 1,230 mm during 2016–17, whereas only 600–996 mm was recorded during the other years. The western disturbances in northwest India caused low rainfall during the winter. During the November–April, 2019–20 and 2014–15 had the highest rainfall, which was 306 mm and 315 mm, respectively, and the 2 years with the least rain, 2015–16 and 2017–18, had 22 and 39 mm, respectively. Most of the rainfall during the study period was received during the monsoon period from July to September (Supplementary Table 1).

As the experiment progressed, the grain yield under PNB and ZT conditions was found to be statistically (p = 0.05) identical. Across years, maize grain yield showed an increase of 4–29% and 3.9–28.7% under the PNB and ZT, respectively, compared to the CT (Supplementary Figure 3a). However, RDF_I and NE_I recorded 20.9–48.6% and 10.3–50.5% higher grain yields than the RDF, respectively (Supplementary Figure 3b). The NE_I consistently produced the highest stover yield and was similar to the RDF_I for five of the 9 years studied (Supplementary Figures 3c,d). ZT and PNB had 12% greater pooled yields than the CT (Figure 2a), but the NE_I and RDF_I stover yields increased by 32 and 26%, respectively, over the RDF (Figure 2b).

During the 2021–22 season, the PNB yielded 4.2 and 21.3% higher mustard yields than the ZT and the CT, respectively (Supplementary Figure 4a). On average, the mustard yield was 29.8 and 27.3% higher under the NE_I and the RDF_I, respectively, than under the RDF (Supplementary Figure 4b). The stalk yield was higher under the ZT and the PNB conditions than under the CT. The NE_I also produced a significantly higher stalk yield than the RDF_I and the RDF (Supplementary Figures 4c,d). Furthermore, the pooled mustard grain and stalk yields were 13.3 and 19.6% higher, respectively, under the PNB over the CT (Figure 2c). The NE_I produced 23.9 and 24.3% higher grain and stalk yields, respectively, over the RDF (Figure 2d).

Long-term tillage significantly influenced the system yields, measured in respect of the maize grain equivalents (EY), in 5 out of the 9 study years, wherein ZT and the PNB improved the EY by 7–20% over the CT (Figure 3a). Regarding the nutrient management practices, the EY was the same between the NE_I and the RDF_I, but being 26.6–30.3% higher than the RDF (Figure 3b). A significant interaction between tillage practices and nutrient management practices was observed in the EY. The highest EY was recorded with the PNB-NE_I during 2013–14, 2018–19, and 2020–21; CT-NE_I during 2014–15; ZT-NE_I during 2015–16, 2017–18, and 2019–20; ZT-RDF_I during 2016–17; and PNB-RDF_I during 2021–22 (Table 2).

The effects of tillage practices on water footprints were significant (p = 0.05), with the lowest water footprint recorded under the PNB (139 liter kg⁻¹ ha⁻¹) closely fb ZT (140.5) liter kg⁻¹ ha⁻¹ and the highest under the CT (156.7) liter kg⁻¹ ha⁻¹ (Supplementary Figure 5). The NE_I (132.4 liter kg⁻¹ ha⁻¹) and RDF_I (135.7 liter kg⁻¹ ha⁻¹) resulted in significantly lower water footprints than the RDF (168.1 liter kg⁻¹ ha⁻¹). On average, the system water productivity was the highest under PNB and fb ZT, which were 12.9 and 11.9% higher than the CT, respectively. The NE_I and the RDF_I-based nutrient management practices improved the system water productivity by 26.6 and 23.4% over the RDF, respectively. There was a significant interaction between tillage and nutrient management practices in terms of water productivity, which followed the same trend as the EY (Table 3).

At tasseling of ninth season maize, sampling was performed from a 0–45 cm soil profile. Among the tillage practices, a significant difference in soil available N was recorded up to 30 cm depth. ZT and PNB had a greater supply of N than CT at 5 cm. However, in 5–15 cm, PNB had a significantly higher N than ZT and CT. Again, in 15–30 m soil depth, PNB was similar to ZT but significantly higher than CT. The soil N did not differ between 30 and 45 cm soil depths due to tillage practices. RDF_I and NE_I indicate more available soil N in 0–30 cm soil depth. However, at 30–45 cm soil depth, NE_I recorded higher soil N than RDF_I and RDF (Figures 4a,b). Across different soil depths, ZT and PNB had significantly higher levels of soil NO₃⁻ N than CT; however, NE_I and RDF_I recorded higher levels of NO₃⁻ N than RDF (Figures 4c,d). In addition, CA-based tillage (ZT/PNB) had higher levels of soil NH₄⁺-N in 0–30 cm soil depth than CT but was similar in 30–45 cm depth. In the top (0–5 cm) and bottom (30–45 cm) soil layers, nutrient management practices recorded similar soil NH₄⁺-N. However, at 5–30 cm soil depth, NE_I and RDF_I were significantly superior to RDF (Figures 4e,f).

ZT had a larger bacterial population (10⁷), similar to PNB, but significantly greater than CT. Again, in nutrient management, NE_I and RDF_I showed much larger bacterial counts than RDF plots. Interestingly, the 5–30 cm soil depth had 8.7 and 38.6% more bacterial populations than the 0–5 cm and 30–45 cm soil depths, as a result of adequate moisture (Figures 5a,b). In addition, these CA-based practices recorded significantly higher fungi (10⁴) and actinomycetes (10⁴) counts than the CT-based. However, NE_I and RDF_I had higher fungal counts than the RDF plots (Figures 5c–f). At 0–5 cm soil depth, PNB had the highest microbial biomass-P (MBP), fb ZT, and CT. At the next soil depths, both ZT and PNB had similar MBPs and were significantly larger than the CT plots. Across soil depths, NE_I and RDF_I-dominated RDF plots (Figures 6a,b). Furthermore, the PNB plots recorded increased soil microbial biomass-C (MBC) at 0–5 cm depth compared to the ZT and CT plots. However, in subsequent layers (5–45 cm), ZT resulted in more MBC activities than PNB and CT plots. NE_I facilitated greater MBC at 0–5 cm soil depth. On the contrary, RDF_I had greater MBC at 5–45 cm soil depth than NE_I and RDF plots (Figures 6c,d).

The maximum contribution to non-renewable inputs in the system was from indirect sources, such as the fertilizers and the pesticides, accounting for ~75% of the total. Direct non-renewable inputs, such as diesel and electricity, were the second largest contributors under the CT. In contrast, indirect renewable inputs, including crop seeds and residue, accounted for ~57% of the inputs used under PNB and ZT conditions. The CT required more inputs, such as land preparation, labor use, irrigation, and electricity, than the ZT and PNB (Supplementary Figures 6a,b). The energy input (E_I) was highest under PNB (110,585 MJ ha⁻¹), fb the ZT (110,173 MJ ha⁻¹), whereas it was lowest under CT (53,036 MJ ha⁻¹). However, the energy output (E_O) was found to be the highest under ZT conditions and was similar to that under PNB conditions, whereas it was the lowest under CT conditions due to the lower system biological yields. The NE_I registered the maximum E_O at 7.6 and 28.7% higher than the RDF_I and RDF, respectively. Furthermore, the net energy (N_E), energy use efficiency (E_UE), and energy productivity (E_P) were found to be higher under the CT than under the PNB and ZT. Similarly, among the nutrient management practices, the N_E, E_UE, E_P, and the non-renewable energy ratio (N_RER) were the highest under the NE_I, followed by the RDF_I and the least under the RDF. The specific energy (S_E) was the highest under the ZTFB (18.2 MJ kg⁻¹) and RDF (17.8 MJ kg⁻¹). The interaction between tillage and nutrient management practices indicated that the maximum E_I and E_O were recorded under PNB-NE_I, N_E, E_UE, and the E_P under the CT-NE_I, N_RER under ZT-NE_I, and S_E under PNB-RDF treatment combinations (Table 4).