Water Management Alters Emission Behaviour: Comparative Analysis of Methane Flux Drivers in Three Rice Systems Using Generalised Additive Models

Alternate Wetting and Drying (AWD) is widely advocated as a strategy to reduce methane (CH₄) emissions from rice paddies by decreasing the duration of flooding. However, AWD implementation can differ substantially across climatic regions and agricultural systems. It is not yet established whether water management affects only the magnitude of CH₄ emissions or also modifies the dominant environmental controls governing sub-daily flux variability, particularly the interaction between water status, temperature, and other micrometeorological variables. 

The current research evaluates whether different water-management regimes yield distinct emission-control outcomes by comparing the nonlinear hierarchies of ecological drivers influencing half-hourly CH₄ fluxes across three rice systems: continuously flooded (CF) and two distinct AWD practices. 

Half-hourly CH₄ fluxes and associated drivers were analysed from three rice systems: the Philippines, Japan, and South Korea. Fluxes were standardised and paired with engineered hydrologic and micrometeorological predictors, including water depth, depth-change rates, hydroperiod integral (hydrologic memory), psychrometric variables, diurnal harmonics, and interaction terms. Multivariate generalised additive models (GAMs) were constructed using normalised predictors and assessed with 80/20 train–validation splits. The importance of each driver was determined using permutation ΔRMSE and drop-one diagnostics. 

Three distinct emission-control regimes were identified. In Japan (characterised by continuous flooding), moderate mean emissions (5.70 mg CH₄ m⁻² h⁻¹) were regulated mainly by water–temperature interactions, suggesting thermal buffering by standing water. South Korea (AWD with regular wet–dry cycling) exhibited the highest emissions (16.71 mg CH₄ m⁻² h⁻¹) and a transition toward direct atmospheric forcing, with air temperature as the dominant predictor and minimal influence from water–temperature interactions. The Philippines (aerobic-dominated AWD) demonstrated the lowest emissions (1.92 mg CH₄ m⁻² h⁻¹), with hydrologic memory and dryness as the primary modulators. 

Water management influences both the magnitude of CH₄ emissions and the dominant controlling mechanisms: thermal buffering prevails under continuous flooding, atmospheric forcing under frequent wet–dry cycling, and hydrologic memory under aerobic-dominated AWD. Our analysis shows that AWD encompasses fundamentally different emission regimes. The climate benefits of AWD depend on drainage depth, cycle frequency, and the persistence of aerobic conditions. An AWD typology that distinguishes practices by their dominant control mechanisms is suggested to strengthen emission inventories, MRV frameworks, and management guidance.