Temporal variations in the amount of radionuclides released into the atmosphere during the Fukushima Dai-ichi Nuclear Power Station (FNPS1) accident and their atmospheric and marine dispersion are essential to evaluate the environmental impacts and resultant radiological doses to the public. In this paper, we estimate a detailed time trend of atmospheric releases during the accident by combining environmental monitoring data with atmospheric model simulations from WSPEEDI-II (Worldwide version of System for Prediction of Environmental Emergency Dose Information), and simulations from the oceanic dispersion model SEA-GEARN-FDM, both developed by the authors. A sophisticated deposition scheme, which deals with dry and fogwater depositions, cloud condensation nuclei (CCN) activation and subsequent wet scavenging due to mixed-phase cloud microphysics (in-cloud scavenging) for radioactive iodine gas (I 2 and CH 3 I) and other particles (CsI, Cs, and Te), was incorporated into WSPEEDI-II to improve the surface deposition calculations. The fallout to the ocean surface calculated by WSPEEDI-II was used as input data for the SEA-GEARN-FDM calculations. Reverse and inverse source-term estimation methods based on coupling the simulations from both models was adopted using air dose rates and concentrations, and sea surface concentrations. The results revealed that the major releases of radionuclides due to FNPS1 accident occurred in the following periods during March 2011: the afternoon of 12 March due to the wet venting and hydrogen explosion at Unit 1, the morning of 13 March after the venting event at Unit 3, midnight of 14 March when the SRV (Safely Relief Valve) at Unit 2 was opened three times, the morning and night of 15 March, and the morning of 16 March. According to the simulation results, the highest radioactive contamination areas around FNPS1 were created from 15 to 16 March by complicated interactions among rainfall, plume movements, and the temporal variation of release rates associated with reactor pressure changes in Units 2 and 3. The modified WSPEEDI-II simulation using the new source term reproduced local and regional patterns of cumulative surface deposition of total 131 I and 137 Cs and air dose rate obtained by airborne surveys. The new source term was also tested using three atmospheric dispersion models (MLDP0, HYSPLIT, and NAME) for regional and global calculations and showed good agreement between calculated and observed air concentration and surface deposition of 137 Cs in East Japan. Moreover, HYSPLIT model using the new source term also reproduced the plume arrivals at several countries abroad showing a good correlation with measured air concentration data. A large part of deposition pattern of total 131 I and 137 Cs in East Japan was explained by in-cloud particulate scavenging. However, for the regional scale contaminated areas, there were large uncertainties due to the overestimation of rainfall amounts and the underestimation of fogwater and drizzle depositions. The computations showed that approximately 27% of 137 Cs discharged from FNPS1 deposited to the land in East Japan, mostly in forest areas.