Sustainable intensification for a larger global rice bowl

Rice provides 21% of global calorie intake using 11% of global cropland. Demand is projected to rise from 480 Mt in 2014 to ~550 Mt by 2030. Yet several sustainability concerns emerge: yield growth is slowing or plateauing in key regions; rice consumes large shares of global irrigation water, fertilizers, and pesticides and contributes substantially to CH4 and N2O emissions; and rising labor costs challenge profitability. Sustainable intensification—producing more on existing cropland with improved efficiency of energy, nutrients, water, and pesticides—offers a pathway. The study asks how to prioritize national-to-global R&D investments by quantifying yield gaps (actual yield relative to biophysical yield potential) and benchmarking resource-use efficiency using yield-scaled metrics across diverse rice systems. Prior efforts are local or regional; a comprehensive global assessment has been lacking. This work evaluates 32 cropping systems in 18 countries (covering 51% of global rice area) to identify where yield can increase, environmental impacts can decrease, or both.

The paper synthesizes evidence that: (i) yield gains have plateaued in several major rice regions (e.g., California, China, Indonesia, South Korea); (ii) rice production has high environmental footprints, including large shares of global irrigation water, fertilizer, pesticide use, and GHG emissions (CH4, N2O); (iii) prior benchmarking of yield gaps and resource-use efficiencies mainly focused on specific countries or regions; and (iv) yield-scaled metrics better capture environmental performance across systems than area-based metrics. Existing studies show high yields can coincide with lower yield-scaled impacts in cereals, but a global, system-level evaluation for rice to inform R&D prioritization was missing.

Scope and systems: 32 rice cropping systems in 18 rice-producing countries representing 51% of global harvested rice area. Systems span tropical and non-tropical climates; irrigated and rainfed lowland (plus one upland system in northern Brazil); and single, double, or triple cropping. Metrics are calculated per crop (averaged over within-year crop cycles) and also annually (in Supplementary Information). Area-weighted averages use harvested area shares. Data sources: Structured questionnaires completed by regional experts collected field size, tillage, establishment method, mechanization level, seed rates and dates, fertilizer and manure use, pesticide applications (products, rates, counts), irrigation amounts and energy source, labor, and straw management. Actual yields (standard moisture 14%) per crop cycle were averaged over at least three recent seasons (some exceptions). Data were cross-validated with FAOSTAT, World Bank, IFA, literature, and experiment-based irrigation estimates. Weather data came from representative stations. Yield potential and yield gaps: Yield potential (irrigated) or water-limited yield potential (rainfed) from the Global Yield Gap Atlas (GYGA) using ORYZA2000/ORYZA v3 (APSIM for India); Australia from Lacy et al. Yield gap is actual yield expressed as % of potential per crop (also annual provided). Normalization by potential enables fair cross-system comparison. Resource-use and environmental metrics: Computed on area and yield-scaled bases: (i) Global Warming Potential (GWP; CO2-eq) including embedded emissions from input production/transport, on-farm fossil fuel (including irrigation pumping), direct and indirect N2O (direct via N surplus approach; indirect = 20% of direct), and CH4 using IPCC methods; steady-state soil C assumed for lowland rice. (ii) Energy inputs (embodied plus direct; strongly correlated with GWP, reported in Supplementary). (iii) Water supply = irrigation + in-season precipitation. (iv) Pesticides: number of applications per crop; Environmental Impact Quotient (EIQ) and active ingredient amounts estimated where possible, but application counts used for cross-system comparison due to data uncertainty. (v) Nitrogen balance = inputs (fertilizer, manure, biological N fixation) − N removal in grain and exported/burned straw; assumed biological N fixation 30 kg N ha−1 crop−1 in lowland and 10% of that in upland; used 75 kg N ha−1 threshold to define excess N. (vi) Labor input: hours per hectare and yield-scaled (h per Mg), with characterization of mechanization, establishment method, and field size. Overall performance index: Six components—yield gap (% of potential), and yield-scaled GWP, water supply, pesticide applications, N balance (as absolute deviation from 8 kg N Mg−1 grain, target based on AU and CA), and labor—were normalized to maxima and aggregated with weights to balance yield gap, resource-use efficiency, and labor. Lower index indicates better performance. Radar charts shown by climate zone. Scenario analysis: Two global scenarios over current areas and cropping intensities: (1) Raise average yields to 75% of potential in 19 systems with current yields <60% of potential and reduce N balance to 75 kg N ha−1 in eight systems with N balance >100 kg N ha−1. (2) Close yield gaps to 75% with unchanged current yield-scaled N balance. Computed changes in rice production and excess reactive N. Statistics: Pearson correlations evaluated associations (significance p<0.01 unless noted).

• Yield potential and realized yields: Average yield potential across systems is 9.5 Mg ha−1 crop−1 (range 5.9–14.8). Non-tropical systems have higher per-crop potential (9.9 vs 8.8 Mg ha−1), but tropical systems have higher annual potential due to greater cropping intensity (15.3 vs 12.2 Mg ha−1 year−1). Average actual yields are 57% of potential, with wide variation: irrigated systems in Egypt, northern China, Australia, and California achieve ~75% of potential, whereas rainfed lowland in Sub-Saharan Africa and upland in northern Brazil reach only 20–40%.
• Global warming potential (GWP): Area-based GWP increases with higher yield (r=0.76), but yield-scaled GWP decreases as yield gap closes (r=−0.60). High-yield systems (e.g., Egypt, N China, Australia, California) have higher per-ha GWP but lower per-Mg GWP than low-input, low-yield systems (e.g., Sub-Saharan Africa), implying less land needed and potentially lower land-use impacts for a given production target.
• Water supply: No significant association between yield gap closure and area-based water supply (p=0.50). Yield-scaled water supply declines with greater yield gap closure (r=−0.72); lowest in well-managed, fully irrigated semiarid systems (California, Egypt, Australia).
• Pesticides: Number of applications per crop increases with yield level (r=0.51), likely reflecting greater pest pressure in denser canopies and high cropping intensity in the tropics. The relation between yield-scaled pesticide use and yield is weaker; excluding Sub-Saharan Africa reveals a negative association (r=−0.48, p<0.05).
• Nitrogen: Yield gap closure correlates with higher N input (r=0.75) and moderately with higher N balance (r=0.39). Systems with small yield gaps often have N inputs >150 kg N ha−1 and N balances >50 kg N ha−1, with yield-scaled N balances 0–20 kg N Mg−1. Some high-performing systems (California, Australia) maintain relatively small N balances (50–75 kg N ha−1; <10 kg N Mg−1), while others (southern USA; southern and central China) show large N balances (>100 kg N ha−1; >15 kg N Mg−1), indicating scope to reduce N without yield loss. Several Sub-Saharan African systems exhibit negative N balances (soil N mining), indicating the need for higher N inputs or improved soil N supply. The yield vs yield-scaled N balance shows a curvilinear relation (r=0.62), with larger yield gaps at both very low and very high yield-scaled N balances.
• Labor: Labor per crop spans >100-fold (7–900 h ha−1). Highly mechanized, large-field, direct-seeded systems (USA, Australia, Uruguay) require <40 h ha−1, whereas less mechanized, small-field, transplanted systems in Sub-Saharan Africa and Asia require >400 h ha−1. Yield-scaled labor declines as yield gap closes (r=−0.72) across both low- and highly mechanized systems (e.g., 1 h Mg−1 in USA/Australia vs ~12 h Mg−1 in South America; >200 h Mg−1 in parts of Sub-Saharan Africa vs ~110 h Mg−1 in SE Asia/China).
• Overall performance: Non-tropical systems generally outperform tropical systems, but top performers exist in both (California, Australia, northern China; Vietnam, Thailand). Several South and South-East Asian systems have disproportionately high yield-scaled pesticide use and N balance.
• Scenario impacts: Raising yields to 75% of potential in 19 low/intermediate-yield systems increases annual production by 146 Mt (+32%), sufficient to meet projected 2030 demand. Concurrently capping N balance at 75 kg N ha−1 in eight high-N systems would reduce excess reactive N by 2 Mt (−95%). If yield gaps close without improving yield-scaled N balance, excess reactive N would instead rise to 5.0 Mt (+140%).

Findings demonstrate that high yields and high resource-use efficiency per unit grain are compatible goals. Strategic R&D should prioritize systems with large yield gaps and/or poor yield-scaled environmental performance to maximize global gains. Where yields already approach 75% of potential (Egypt, California, Australia, China), substantial gains are limited; breeding to raise potential and fine-tuning management may provide incremental improvements while further reducing yield-scaled impacts. In contrast, systems in Sub-Saharan Africa and some South-East Asian countries (e.g., Philippines, Myanmar) can benefit from greater and better-managed inputs (fertilizers, pesticides) coupled with robust extension and agronomy to minimize environmental externalities. Many South and South-East Asian systems appear able to increase yields while reducing N surpluses via knowledge-intensive, site-specific nutrient and water management. Rainfed lowland systems have large yield gaps and higher yield-scaled water use due to risk exposure; risk mitigation (e.g., supplemental irrigation, insurance) is crucial. Socioeconomic factors strongly influence outcomes: yield gap closure correlates with GDP per capita (r=0.68), reflecting access to finance, inputs, mechanization, markets, and extension. Interventions must consider local trade-offs (e.g., water-saving practices may elevate risk or require sophisticated management) and be tailored to biophysical and socioeconomic contexts.

This global benchmarking across 32 rice cropping systems shows substantial scope to raise production and reduce environmental impacts through targeted sustainable intensification. Focusing R&D and policies on systems with large yield gaps and/or excessive N balances could add ~146 Mt of annual rice (+32%) while nearly eliminating excess reactive N if N balances are constrained to ~75 kg N ha−1. High-yield systems can continue to lower yield-scaled impacts via refined management and improved cultivars. Future work should: (i) include data-poor but important systems (e.g., drought-prone rainfed lowlands in NE Thailand and eastern India); (ii) increase spatial granularity of yield, management, and emissions data; (iii) integrate watershed-scale water sustainability; and (iv) couple agronomic interventions with enabling institutions and policies to support adoption, risk mitigation, and environmental safeguards.

Uncertainties arise from modeled yield potentials, survey-based management and yield data, and assumed emission factors. Some systems (e.g., rainfed lowlands in NE Thailand and eastern India) lacked robust data and were excluded, limiting generalizability to those environments. Spatial aggregation was necessary for cross-system comparisons and may mask within-system variability. GWP estimates assumed steady-state soil C in lowland rice and did not estimate soil C changes for upland Brazil. Pesticide risk comparisons relied on application counts due to uncertain EIQ inputs. Despite these limitations, cross-validation with independent datasets and literature supports the robustness of main conclusions.