Research progress in crop yield estimation models based on spectral reflectance data

Accurate and timely estimation or prediction of crop yield at a regional scale is critical for many applications such as food security warning system, agricultural land management, food trade policy and carbon cycle research. Remotely sensed data can provide the spatial and temporal information of land surface at various scales, and are an attractive tool for assessment of the magnitude and variation of crop condition parameters. Thereby, spectral reflectance data (400～2500 nm) have become the primary source for crop yield estimation at a regional level. This paper introduces the theory and basic principles of crop spectral reflectance and the impact factors for crop yield, firstly. Then, it presents an overview of different methods that can be used to estimate crop yield at a regional scale from optical remote sensing data in the VNIR-SWIR regions. The overview is based on an extensive literature review, where three methods are discussed: empirical methods, semi-empirical methods, and mechanistic methods. Among these methods, the assimilation of reflective remote sensing data into crop growth model is thought to be of increasing importance for crop yield assessment. The studied literature reveals that many valuable models have been developed for the crop yield estimation by using spectral reflectance data. However, for improving the estimation accuracy and developing operational system in the near future, emphasis must be put on deriving biophysical and biochemical canopy state variables by statistical-empirical and physically based approaches, and on the fusion of remote sensing data with different spatial, temporal, spectral and angular resolutions, and on assimilating remotely sensed data into crop growth model.