The field experiment was carried out during 2021–2022 and 2022–2023 at ICAR-National Institute of Abiotic Stress Management (NIASM), India (18°09′30.62″ N latitude and 74°30′30.08″ E longitude, altitude of 570 m above mean sea level) ( Figure 1 ). The site falls within the hot semi-arid agro-ecological region in the Deccan Plateau of India, known for its extremely high temperatures, unpredictable rainfall patterns, and extended periods of dry weather (Pradhan et al., 2023). Its long-term average annual rainfall is 576 mm, 70% of which occurs during June–September as southwest monsoons and 21% during October–December as retreating monsoons. The average values for temperature (maximum and minimum), relative humidity (maximum and minimum), and total rainfall and open pan evaporation during the crop growing periods (November–March) were 30.8°C and 15.7°C, 84% and 39%, and 8.2 and 23.0 mm during 2021–2022 and 31.5°C and 14.3°C, 79.9% and 30.4%, and 7.8 mm and 25.1 mm during 2022–2023, respectively (Automatic Weather Station, ICAR-NIASM). The details of the weather parameters are provided in Supplementary Table 1 . The soil of the experimental site is originated from parental basaltic rocks and characterized as shallow murrum (up to 40-cm depth) with 30% stones (>2 mm): 69% sand, 21% silt, 10% clay, and 5% available moisture holding capacity. At the beginning of the experiment, the pH (1:2.5 soil:water), electrical conductivity (EC), Walkley–Black carbon (C), KMnO₄ oxidizable nitrogen (N), 0.5 M NaHCO₃ extractable phosphorous (P), and 1 N NH₄OAc extractable potassium (K) were 7.2, 0.18 dS m⁻¹, 0.14%, 98.32 kg ha⁻¹, 2.51 kg ha⁻¹, and 207 kg ha⁻¹, respectively.

The experiment was laid out in a split-split plot design with four sowing dates (D₁, November 1; D₂, November 15; D₃, December 1, D₄, December 15) in main plots, two irrigation practices (I₁, 40% ET_c; I₂, 80% ET_c) in sub plots, and three nitrogen doses (N₁, 100 kg N ha⁻¹; N₂, 150 kg N ha⁻¹; N₃, 200 kg N ha⁻¹) in sub-sub plots having three replications ( Supplementary Figures 1 , 2 ). The nitrogen quantity was applied in two splits: 50% as basal and 50% during the flowering period. The P and K doses (i.e., 60 kg ha⁻¹) were applied as basal in all the treatments. The experiment was initiated in November 2021 using the quinoa accession "Jaipur Local", manually sown at a seed rate @ 5 kg ha⁻¹ with 45 cm × 15 cm spacing. Sowing was established by dibbling five to six seeds per pocket in the soil to a depth of 1–2 cm. The gross and net plot areas under each treatment were 9 × 5 m and 7.5 × 4 m, respectively. To avoid border and interaction effects, a buffer area of 1.5 m was left between experimental units. Irrigation to the crops was provided via surface drip irrigation (having a discharge rate of 4 L h⁻¹ through 16-mm laterals having inline emitters), and the crops were irrigated based on the actual crop evapotranspiration (ET_c) approach. The crop ET_c was computed using the following equation (Equation 1):

where K_p is the pan coefficient (0.70), E_pan is the cumulative pan evaporation (mm), and K_c is the crop-specific coefficient. The K_c values for quinoa as reported by Garcia et al. (2003) were considered in our study. One common irrigation of 30 mm was provided after sowing for uniform germination and crop stand establishment. Scheduling of later irrigations was conducted as per the treatment, i.e., at 40% and 80% ET_c to quinoa. The total quantity of water applied through drip irrigation to the cropping systems in both the study years is given in Supplementary Table 2 . All recommended crop-specific packages and practices of weeding, intercultural operations, and disease pest management were strictly followed.

When the seeds matured and plants started drying, all the morphological and yield attributing parameters were measured from a total of 10 plants per treatment. The shoots (only stem and leaves) and roots were oven-dried at 60°C for 48 h to determine the respective dry matter content. The crops were manually harvested using sickles as they matured, followed by sun drying for 7 days. The dried panicles for each experimental plot were threshed and winnowed manually. Seed yield was recorded at 12% moisture content, while the stover was oven-dried at 60°C until constant weight was obtained and then expressed as kg ha⁻¹ for respective treatments.

Water productivity (WP) was calculated using the following equation (Equation 2).

The phenological developments, viz., 50% germination, 50% visible bud, 50% flowering, and 50% maturity were recorded as and when at least 50% of the plants were showing the indications in the whole plot. Days to maturity was calculated from the date of emergence to the date when the crop was harvested.

Growing degree days (GDD) was calculated based on the following formula (Equation 3):

where T_max is the maximum temperature, T_min is the minimum temperature, and T_base is the base temperature, which was 3°C for quinoa (Jacobsen and Bach, 1998).

The seed protein content was determined using the Kjeldahl method of N estimation from plant samples with a conversion factor of 6.25. The seed protein content was then multiplied with the seed yield to estimate the protein yield under corresponding treatments. The total saponin content of quinoa seed and husk was estimated following the colorimetric determination procedure reported by Hiai et al. (1976). All quinoa samples were ground with a blender and passed through 1-mm sieve. Then, 10 g of each sample powder was dissolved in 40 mL of 25% ethanol and kept in a mechanical shaker for 12 hrs. Then, 0.5 mL of vanillin solution was added to 0.5 mL of aqueous ethanol sample, followed by the addition of 5 mL of 72% sulfuric acid and mixed in an ice-water bath. The mixture was then warmed in the bath at 60°C for 10 minutes, followed by cooling in an ice-water bath. A water blank with the reagents was also made. Absorbance at 450 nm was recorded against the blank with the reagents using a spectrophotometer. Quillaja saponin was used as a standard, and the total saponin content was expressed as g 100 g⁻¹ of dry weight.

The total C input and output were determined by adding the carbon equivalents of all inputs and outputs during crop production. Carbon equivalents (CE) of all inputs, operational activities, and processes were combined to estimate the source-wise contribution to C input ( Supplementary Figure 3 ; Supplementary Table 3 ). Similarly, the C output of each cropping system was quantified using methodologies provided by Choudhary et al. (2017) and Equation 4.

considering that C content accounts for 44% of the total plant biomass as given by Lal (2004).

Carbon efficiency, an indicator of total C production over the total input C, was calculated as follows (Equation 5).

Carbon sustainability index (CSI) was estimated as the net gain in C over the total C input as given by Lal (2004) and Choudhary et al. (2017) and depicted in Equation 6.

Carbon footprint (CF): Carbon footprints of the cropping systems indicated the total greenhouse gas (GHG) emissions in terms of kg CE ha⁻¹ to produce 1 kg of the economic product and was computed as per Equation 7 as suggested by Yadav et al. (2021).

Estimation of nitrous oxide (N₂O) emission included both direct and indirect emissions. As direct N₂O emission is proportional to the amount of N applied, direct N₂O emission was computed using Equation 8.

where 0.016 is the default emission factor for N fertilizer application, 1.571 is the conversion factor used to convert N₂ to N₂O, and 298 is the global warming potential (GWP) of N₂O concerning CO₂.

Similarly, indirect N₂O emissions included loss of N fertilizer in the form of volatilization and were calculated using the Intergovernmental Panel on Climate Change (IPCC) guidelines of Tier 1 and Equation 9.

where 0.1 is the fraction used for volatilization, 0.010 is the default emission factor used for volatilization, 1.571 is the conversion factor used to convert N₂ to N₂O, and 298 is the GWP of N₂O concerning CO₂.

The recorded data were statistically analyzed using analysis of variance (ANOVA) for split-split plot design using the "Agricolae" package of R (R Development Core Team, 2015). A mixed model was used considering sowing date as the main plot factor, irrigation as sub plot factor, and nitrogen management as the sub-sub plot factor and analyzed in a split-split plot design. Since no major differences were observed among the treatments for the recorded observations, the results were averaged for 2021–2022 and 2022–2023. The F-test and least significant difference (LSD) (p < 0.05) were used to decipher the significance of the means of treatments and their interactions.

The impact of different sowing dates on quinoa's growth duration, developmental stages, and corresponding temperature ranges is illustrated in Figure 2 . For sowing on November 1, the vegetative stage continued for 38 days after sowing (DAS), with temperatures ranging from a maximum of 27.35°C to 31.45°C and a minimum of 12.45°C to 19.1°C. The flowering period extended over 22 days, with temperatures reaching a maximum of 27.9°C to 31.4°C and a minimum of 11.25°C to 18.2°C. The seed-filling and maturity stages began at 60 DAS and continued for 33 days, resulting in a total crop duration of 99 days. During this phase, maximum temperatures ranged from 28°C to 31.5°C, while minimum temperatures ranged from 10.35°C to 14.9°C. Sowing on November 15 reduced the total crop duration by 1 week compared to November 1. The vegetative stage lasted 35 days, with a wider window for flowering (30 days). The seed-filling and maturity stages started at 65 DAS and lasted for 33 days. Temperatures during these stages were as follows: vegetative stage (27.35°C to 31.8°C maximum and 11.25°C to 19.1°C minimum), flowering stage (28°C to 31.4°C maximum and 11°C to 16.05°C minimum), and seed-filling and maturity stages (27.15°C to 32.1°C maximum and 10.35°C to 14.5°C minimum), respectively. Sowing in December resulted in a longer flowering period (35–40 days) and shorter seed-filling and maturity periods (20–23 days). During the seed-filling and maturity phases, temperatures were notably higher: 31.3°C to 34.85°C maximum and 11.6°C to 15.25°C minimum for December 1 sowing and 31.25°C to 35.25°C maximum and 14.5°C to 19.15°C minimum for December 15 sowing.

For cumulative growing degree days (CGDD), the highest accumulation was observed for the earliest sowing date, November 1, with 1,900.40°C ( Table 1 ). Sowing on November 15, December 1, and December 15 resulted in lower CGDD values of 1,785.30°C, 1,661.75°C, and 1,789.25°C, respectively.

Sowing date and nitrogen levels showed significant effects (p < 0.05) on quinoa height, dry matter accumulation, and yield attributes ( Table 2 ). The maximum plant height (113.67 cm) was obtained during the November 1 sowing. Height decreased by 7%–20% with later sowing dates, reaching a minimum (91.28 cm) for sowing on December 15. A similar trend was seen for both shoot and root dry matter, ranging from 11.39 to 72.43 g plant⁻¹ and 3.52 to 8.69 g plant⁻¹, respectively. Among the yield attributes, sowing on November 1 produced the maximum panicle weight, panicle length, seed weight, and husk weight, which were 2.6, 1.7, 7.2, and 2.7 times, respectively, than those from December 15 sowing (19.05 g plant⁻¹, 18.24 cm plant⁻¹, 10.15 g plant⁻¹, and 4.70 g plant⁻¹). However, irrigation levels did not significantly affect these parameters except panicle length, which was 8.8% higher in 80% ET_c than in 40% ET_c (22.93 cm plant⁻¹) ( Table 2 ). Among the nitrogen doses, the highest mean values for plant height, shoot dry matter, and yield attributes were achieved with N₃ (200 kg N ha⁻¹) and the lowest with N₁ (100 kg N ha⁻¹). The application of 150 kg N ha⁻¹ was significant as compared to 100 kg N ha⁻¹ only for shoot dry matter and seed weight. The maximum seed weight was observed under N₃, which was 33% higher than that under N₂ (39.25 g plant⁻¹) and 56% higher than that under N₁ (33.40 g plant⁻¹). The seed weight of N₂ was 18% higher than that of N₁ (p < 0.05). Husk weight was significantly higher for November sowing (12. 8 g plant⁻¹) and for 200 kg N ha⁻¹ (11.95 g plant⁻¹) (p < 0.05). The 1000-seed weight of quinoa ranged from 2.22 to 2.74 g, with plots sown in November recording a 17% higher value than that of December sown plots (2.2 g) (p < 0.05). Similarly, providing irrigation at 80% ET_c and 200 kg N ha⁻¹ reported 10% and 7% higher 1000-seed weight than 40% ET_c (2.35 g) and 100 kg N ha⁻¹ (2.39 g), respectively.

The seed yield was significantly higher in plots sown on November 1 (1,446 kg ha⁻¹), which was reduced by 50% for sowings on November 15 and December 1 sowing, and the lowest yield was recorded for December 15 sowing (345.61 kg ha⁻¹) (p < 0.05) ( Figure 3 ). There were no differences in seed yield between the two irrigation levels. However, nitrogen levels had a significant effect on quinoa seed yield (p < 0.05) with maximum values under N₃ (916.58 kg ha⁻¹) and N₂ (815.04 kg ha⁻¹), with the latter being at par with N₁ (723.50 kg ha⁻¹).

The water productivity of quinoa production ranged from 0.85 kg m⁻³ to 0.18 kg m⁻³ for our study ( Figure 4 ). Water productivity was maximum for November 1 sowing (0.85 kg m⁻³), followed by November 15 (0.52 kg m⁻³) and December 1 (0.37 kg m⁻³), and minimum for December 15 (0.18 kg m⁻³). Providing irrigation at 40% ET_c was 69.38% higher water productive than that at 80% ET_c (0.36 kg m⁻³). However, water productivity was not significantly influenced by the nitrogen levels.

Since the treatment had no significant effect on seed protein content, the system protein yield followed a similar trend as that of seed yield ( Supplementary Table 4 ; Supplementray Figure 4 ). Seed and husk saponin contents evaluated in the current study ranged from 0.51 to 1.26 g per 100-g dry weight and 0.64 to 1.67 g per 100-g dry weight, respectively ( Figure 5 ). The seed saponin content was 19% higher in November sown crops than in the December 1 sowing (1.01 g 100 g⁻¹ dry weight). However, the minimum saponin content of 0.51 g 100 g⁻¹ dry weight was reported for the December 15 sowing. Similarly, providing irrigation at 40% ET_c had 39% lower saponin as compared to that under 80% ET_c (1.26 g 100 g⁻¹ dry weight). The application of a higher dose of N increased the seed saponin content with N₁ having the lowest value (0.77 g 100 g⁻¹ dry weight), followed by N₂ (1.06 g 100 g⁻¹ dry weight) and N₃ (1.21 g 100 g⁻¹ dry weight). The saponin content of husk was higher than that of seed with irrigation levels having a significant impact, following a similar trend to that of seed saponin content. Delayed sowing, i.e., on December 15, also reported a reduced husk saponin content than the rest of the sowing dates.

Averaged over 2 years, the carbon budgeting and related indices differed among the treatments ( Table 3 ). Early sowing dates, i.e., November 1 sowing (D₁), had the lowest C footprint (0.19 kg CE kg⁻¹ seed) and the highest C efficiency (3.69) and were more sustainable (CSI, 2.69) (p < 0.05). This may be due to improved biomass production leading to higher C output (1,018.26 kg CE ha⁻¹) and moderate C input (276 kg CE ha⁻¹) in S₁. Sowing during December 15 (D₄) resulted in the highest C footprint (0.80 kg CE kg⁻¹ seed) and the lowest C efficiency (0.88) and was not sustainable (CSI, −0.12), which was due to significantly less C output (243.31 kg CE ha⁻¹) as proportionate to the quantity of C input (279.53 kg CE ha⁻¹). Similarly, irrigating the crop at 40% ET_c proved to be more C efficient and sustainable as compared to irrigating them at 80% Etc. Among the nitrogen levels, the application of 100 kg N ha⁻¹ (N₁) registered the lowest C footprint (0.33), higher C efficiency (2.14), and CSI (1.14), which was comparable to N₂, i.e., 150 kg N ha⁻¹. However, higher N levels, i.e., application at 200 kg ha⁻¹ (N₃), resulted in higher C output (632.07 kg CE ha⁻¹) but at the cost of efficiency and increased C footprint; therefore, they were less sustainable (CSI, 0.86).

Among the treatments, N₂O emissions (both direct and indirect) were significantly influenced only by N levels ( Figure 6 ). The N₂O emissions increased with the applied N fertilizer dose. The application of 200 kg N ha⁻¹ had maximum direct (1,497.15 kg CO₂-eq ha⁻¹), indirect (93.57 kg CO₂-eq ha⁻¹), and total (1,590.72 kg CO₂-eq ha⁻¹) N₂O emissions, followed by 150 and 100 kg N ha⁻¹ (p < 0.05).