Nitrous oxide (N₂O) emissions response curves for crops grown outside temperate regions have been rare and have thus far arrived at conflicting conclusions. Most studies reporting N₂O emissions from tropical cropping systems have examined only one or two nitrogen fertilizer application rate(s) which precludes the possibility of discovering nonlinear changes in emission factors (EF, % of added N converted to N₂O-N) with increasing fertilizer-N rates. To examine the relationship between N rates and N₂O fluxes in a tropical region, we compared farming practices with three or four N rates for their yield-scaled impacts from three crops in peninsular India. We measured N₂O fluxes during nine seasons between 2012 and 2015, with N application rates ranging between 0 and 70, 0 and 90, and 0 and 480 kg-N ha^-1 for foxtail-millet (Setaria italica L., locally called korra), groundnut (Arachis hypogaea L., also called peanut) and finger-millet (Eleusine coracana L., locally called ragi), respectively. In two cases, the highest N application rate greatly exceeded crop-N needs. Potential climate smart farming agricultural practices (with low/optimized N rates) led to a 50-150% reduction in N₂O emissions intensity (per unit yield) along with a reduction of 0.2-0.75 tCO2e ha^-1 season^­­-1 as compared to high N conventional applications. We found a non-linear increase in N₂O flux in response to increasing applied N for both N-fixing and non N-fixing crops and the extent of super-linearity for non N-fixing crops was much higher than what has been reported earlier. If a linear fit is imposed on our datasets, the emission factors (EFs) for finger-millet and groundnut were ~3.5% and ~1.8%, respectively. Our data shows that for low-N tropical cropping systems, even when they have low soil carbon content, increase in N use to levels just above crop needs to enhance productivity might lead to relatively small increase in N₂O emissions as compared to the impact of equivalent changes in fertilizer-N use in systems fertilized far beyond crop N needs.

In general, the soils in the two agro-ecological regions are not amenable to cultivation without ploughing. For groundnut and foxtail-millet, tillage was done once in each season about 25 days before sowing. For finger-millet, tillage was done 2-4 times between March and July soon after rainfall depending on soil hardness and manure (if any) was incorporated during the last 1-2 tillage events. Bullock cart ploughing tills soil to a depth of 12 cm and local tractors (used only when the soil is very hard) plough to the depth of up to 18 cm. There was no tillage done to control weeds and there was no use of herbicides and pesticides.

During the rainfed south-west monsoon season (from July to December; locally called kharif), sowing was done manually at a seed rate 146 ± 27 kg ha^-1 for groundnut (Kadiri 6 variety) at a 30 cm row spacing, 10 cm plant spacing and to a depth of 5 cm, 12 kg ha^-1 for foxtail millet (local variety called Jadda Korra) at a 30 cm row spacing, 8-12 cm plant spacing and to a depth of 3-6 cm and 24.7 kg ha^-1 for finger-millet (MR1 variety) at a 25 row spacing to a depth of 3-6 cm. Both millets are sown with a seed drill attached to a bullock and the plots are thinned/weeded 12-20 and 20-25 days after sowing of finger- and foxtail-millet, respectively. The seed rates used in a given crop and season were the same for all treatments. All of the aboveground biomass (as well as belowground biomass for groundnut) was harvested manually 110-130 days after sowing (see exact dates in S1 Table).

N₂O flux monitoring

Manual closed chambers were used to collect air samples from each of the three replicate treatment plots and the air samples were analyzed by electron capture detector (ECD) in a gas chromatograph (Thermo Fisher Trace GC 600) to quantify N₂O emissions rates based on methodology developed in our labs. Because most N₂O emissions occur within 1-4 days following N addition and/or irrigation/rainfall, N₂O flux measurements are more reliable when the sampling frequency is high and the sampling schedule captures spatio-temporal variability in emissions. We performed sampling on 34-60% of the total days in each season (S1 Table), with continuous sampling for 3-5 days after all “events” e.g. sowing, fertilizer application, irrigation/rainfall and weeding with an average of three measurements every week. Stackable manual chambers (size: 30*30*40 cm) were deployed on base-frames. A second 30*30*40 cm chamber was stacked on top of the first chamber if (and only when) the plant height exceeded 40 cm (Fig 2). On each sampling day, four air samples (60 ml each) were collected at 10 minute intervals for 30 minutes between 10 AM and 12 noon to calculate the hourly N₂O flux. The GC was calibrated daily with four standards: 0.197, 0.393, 0.795 and 1.615 ppmv N₂O (Bhuruka Gases, Bengaluru; NIST certified at 2% RSD).  For chamber deployment period of half hour, four sampling points under linear regression, minimum detectable N₂O flux was 33.8 ppb which translates into ~20 µg m^-2 h^-1 for our chambers with a volume of ~36 L, ambient temperatures in the range of 35-45°C and baseframe footprint of 0.09m². Following the recommendation of Parkin and Venterea (2010), we used the actual measured value even if it falls below the minimum detection limit (MDL). The details of the design of chambers and base-frames, methods employed to achieve uniform mixing of headspace air, chamber volume and temperature corrections,  sample storage, data analysis, treatment of negative emissions, calculation of seasonal fluxes and curve fitting have been described previously. Cumulative emissions were calculated separately for each replicate plot before calculating the average emissions for each treatment. For a given farm, when available, the results from different years were averaged for treatments with similar N input rates (Table 1 and S2 Table) to perform multiple regression. The cumulative N₂O flux during a cropping season was calculated by linear interpolation as explained earlier.

Crop yield, yield scaled GHG flux and N mineralization rates

 Yields were measured from each treatment at maturity at the end of a season after separating groundnut pods or millet grains from the plant/straw and sun drying to a constant weight. It is customary in this region to use crop residue to feed cattle and not return it to the plots. Yield scaled GHG flux (i.e., GHGI) for each treatment was calculated by converting average N₂O flux into CO₂e after multiplying by 298 (Global warming potential of N₂O) and dividing by average yield for that treatment. The errors (SE) associated with GHGI were calculated by standard error propagation method. Mineralization rates for the applied organic nitrogen were estimated using methods described earlier. For more details, please look at the submitted pre-print.

No specific programs or softwares are needed to open the data files. The preprint provided with this submission will be available as a technical white paper on the website of Environmenetal Defense Fund.