Microbial production in anoxic wetland rice soils is a major source of atmospheric CH4, the most important non-CO2 greenhouse gas. Much higher CH4 emissions from well managed irrigated rice fields in the wet than in the dry season could not be explained by seasonal differences in temperature. We hypothesized that high CH4 emissions in the wet season are caused by low grain to biomass ratios. In a screenhouse experiment, removing spikelets to reduce the plants' capacity to store photosynthetically fixed C in grains increased CH4 emissions, presumably via extra C inputs to the soil. Unfavorable conditions for spikelet formation in the wet season may similarly explain high methane emissions. The observed relationship between reduced grain filling and CH4 emission provides opportunities to mitigate CH4 emissions by optimizing rice productivity.greenhouse gas ͉ C allocation in rice plants ͉ rice production ͉ global warming A tmospheric methane (CH 4 ) is the second most important greenhouse gas, next to CO 2 . The total global CH 4 emission source is about 600 Tg⅐year Ϫ1 [Tg, teragrams (10 12 g)] and emission of 1 kg of CH 4 to the present-day atmosphere is 21 times more effective in perturbing the radiation balance than emission of 1 kg of CO 2 (1). Wetland rice fields are among the important sources of atmospheric CH 4 , with an estimated source strength of 60 Ϯ 40 Tg⅐year Ϫ1 (2). CH 4 emissions from rice fields have been measured over the past decades in Italy (3), Japan (4), the United States (5-7), China (8), and the Philippines (9). These data have been used in simulation models to predict CH 4 emissions from rice fields, using a limited number of input variables (10, 11). Such predictive models can only help to improve estimates of the global CH 4 source strength if all important processes that control CH 4 emission from rice paddies are incorporated. The Interregional Research Program on Methane Emissions from Rice Fields (1993-1998) (12) provided the first emission measurements for major rice-growing regions by using a standardized automated closed chamber system. One of the most comprehensive data sets from this program is by Corton et al. (13), who measured CH 4 emissions from intensively managed irrigated fields with high-yielding rice, fertilized with N (urea), P, and K. The experiment encompassed nine seasons (five dry seasons and four wet seasons). In reference treatments where rice variety, fertilizer inputs, water management, and organic amendments were kept constant in the successive dry and wet seasons, CH 4 emissions were invariably 1.5-4-fold higher in the wet season than in the dry season (Table 1). We could not link seasonal differences in emission to effects of temperature on bacterial methane production: daily mean and maximum temperatures were very similar in wet and dry seasons, while all other known variables driving CH 4 emission were kept constant. Mean solar radiation from 1994 to 1998 did not differ significantly between the wet seasons (20.6 Ϯ 3.1 MJ⅐m Ϫ2 ⅐day Ϫ1) and dry sea...