This study uses multiple satellite datasets to map paddy rice areas and yields for the Thai Binh Province, Viet Nam, over the summer growing season of 2015. The major datasets used are: first, surface reflectance and vegetation indices (VI) by fusing the optical observations from the Landsat sensors and the MODerate Resolution Imaging Spectroradiometer; and second, the L-band radar data from the PALSAR-2 sensor onboard the Advanced Land Observing Satellite 2. We find that although the fused VI time series are not necessarily beneficial for paddy rice mapping, the fusion datasets reduce observational gaps and allow us to better identify peak VI values and derive their empirical relationships with crop-cutting yield data (R 2 = 0.4 for all the rice types, and R 2 = 0.69 for the dominant rice type -58% of all the sampled fields). The L-band radar data have slightly lower performance in rice mapping than the optical satellite data, while it has much less contribution to yield estimation than the optical data. Furthermore, our study suggests the geolocation errors of satellite images be taken into account when selecting small sample are as for crop cutting. This practice will ensure the representativeness of crop-cutting sample areas with regard to satellite observations and thus better linkages between field data and satellite pixels for yield modeling. We also highlight the need of crop-cutting data from multiple years and/or at different regions to account for the spatial and temporal variations of harvest index to improve the spatially explicit rice yield estimates through satellite observations.