Mitigating nitrogen losses with almost no crop yield penalty during extremely wet years

Abstract

Climate change-induced precipitation anomalies during extremely wet years (EWYs) result in substantial nitrogen losses to aquatic ecosystems (N_w). Still, the extent and drivers of these losses, and effective mitigation strategies have remained unclear. By integrating global datasets with well-established crop modeling and machine learning techniques, we reveal notable increases in N_w, ranging from 22 to 56%, during historical EWYs. These pulses are projected to amplify under the SSP126 (SSP370) scenario to 29 to 80% (61 to 120%) due to the projected increases in EWYs and higher nitrogen input. We identify the relative precipitation difference between two consecutive years (diffPr) as the primary driver of extreme N_w. This finding forms the basis of the CLimate Extreme Adaptive Nitrogen Strategy (CLEANS), which scales down nitrogen input adaptively to diffPr, leading to a substantial reduction in extreme N_w with nearly zero yield penalty. Our results have important implications for global environmental sustainability and while safeguarding food security.

Fig. 1.. Relative changes in nitrogen losses (N_w) under different precipitation anomalies.

Bar plots show N_w changes relative to the 1981–2010 average for maize (A and B), rice (C and D), and wheat (E and F) presented separately for irrigation (left) and rainfed (right) conditions. Error bars represent 95% confidence intervals estimated from 1000 bootstrapped resampled sets. Numbers in blue indicate the average N_w changes with precipitation anomalies >2σ (i.e., extreme N_w changes). The number of FPU-year combinations is indicated by “n.”

Fig. 2.. Cumulative extreme nitrogen losses (N_w).

Each line shows the cumulative extreme N_w relative to the 1981–2010 average along different FPUs. Extreme N_w is set to 0 for the regions without EWYs. Bold red lines indicate the extreme N_w during the historical period, while bold black lines present average extreme N_w among different climate models under future (2036–2065) SSP126 (dashed lines) and SSP370 (solid lines) conditions. Small vertical bars along the bold lines indicate the number of years falling in EWYs. Irrigated (left) and rainfed (right) conditions are distinguished for maize (A and B), rice (C and D), and wheat (E and F).

Fig. 3.. Changes in extreme nitrogen losses (N_w) during 2036–2065 under the scenario SSP370.

Maps show the differences (%) in extreme N_w averaged over five GCMs relative to the 1981–2010 average. Irrigated (left) and rainfed (right) conditions are distinguished for maize (A and B), rice (C and D), and wheat (E and F).

Fig. 4.. Correlation coefficient (r) between extreme nitrogen losses (N_w) and different driving factors of the three crops under irrigated (IRR) and rainfed (RFD) conditions.

The factors include annual precipitation (aPr), N input (N_in), soil bulk density (BD), relative precipitation differences between two consecutive years (diffPr), growing season precipitation (gsPr), precipitation during N fertilization period (ferPr), annual temperature (aT), growing season temperature (gsT), crop yield, coarse fragment (CF), sand content (SDC), and silt content (STC). The asterisk indicates significance at 95%. Details of the correlation refer to figs. S11 to S13.

Fig. 5.. Optimal combinations of diffPr thresholds and scaling ratios of nitrogen input to reduce extreme nitrogen losses (N_w).

The maps illustrate, for each FPU, the best ratio for scaling down nitrogen input (N_in) during years with a relative precipitation difference between two consecutive years (diffPr) higher than a certain threshold under irrigated (left) and rainfed (right) conditions for maize (A and B), rice (C and D), and wheat (E and F). These ratios result in the greatest reduction in extreme N_w while maintaining a <15% reduction in long-term N_in and a <3% reduction in crop yield under the SSP370 scenario. Regions shown in black indicate no optimal diffPr thresholds, and N_in scaling ratios were identified for the given reductions in N_in and yield constraints.

Fig. 6.. Mitigation of future extreme nitrogen losses (N_w).

The bars indicate changes in future extreme N_w, average nitrogen input (N_in), and crop yield for the period 2036–2065, with and without mitigation measures, relative to the 1981–2010 averages. The mitigation measures are achieved by using the diffPr thresholds and N_in scaling ratios shown in Fig. 5. Irrigated (left) and rainfed (right) conditions are distinguished for maize (A and B), rice (C and D), and wheat (E and F).