Projection of pixel-based rice yields by linking geostationary satellite imagery and a crop model in Northeast Asia

An approach of linking satellite imagery and a crop model can allow reproducing geospatial variation in crop productivity efficiently and conveniently. This methodology could be a useful means of responding to ever-increasing food crop demand and managing its production. This study aims to simulate rice yields with geographical changes based on the Communication, Ocean, and Meteorological Satellite (COMS) data incorporated into the GRAMI-rice model in most continental areas of Northeast Asia of interest from 2011 to 2014. The COMS consists of two imager payloads, the Geostationary Ocean Color Imager (GOCI) and the Meteorology Imager (MI), from which surface reflectance and solar radiation were obtained to use as input variables of the model. The GOCI imagery has an advantage of reducing a cloud effect in comparison with any other images from polar orbit satellites, owing to being obtained through frequent observations, i.e., eight times a day. We also used Local Data Assimilation and Prediction System (LDAPS) and Shuttle Radar Topography Mission (STRM) Digital Elevation Model (DEM) data to acquire air temperature and geographic information such as an elevation and a surface slope. Before simulating rice yields, we first performed a classification of paddy fields and estimation of transplanting dates using time-series of spectral indices and geographical characteristics of rice cultivation. After that, the model evaluation that can reflect regional characteristics of the rice cultivars and farming practices was performed and compared with county-level statistical yield data of the study areas. The overall accuracy and the Kappa coefficient of the classified paddy fields were 78.8 % and 51.2 %, respectively. The root mean square error (RMSE), Nash-Sutcliffe efficiencies (NSE), and p-values of two-sample t-tests between observed and simulate rice yields ranged from 0.673 to 0.767 t ha^-1, 0.108 to 0.478, and 0.130 to 0.894, respectively.