Estimating Yield of Canola Based on Time Series of Remote Sensing Data

Spectral vegetation indices have been used as a useful tool in remote sensing to estimate the yield of agricultural crops. However, one factor, which reduces the capability of indices for crop yield estimation, is the limited number of available satellite images. Furthermore, in cases when there are not enough Landsat images, the capabilities of spectral indices in yield estimation using a fusion of MODIS and Landsat data, have been less investigated. The aim of this paper is, first, to introduce the most efficient index/indices for estimating the canola yield and, second, to try to use data fusion techniques in order to increase the efficiency of the selected index/indices. Due to flowering in the growth period, canola has special spectral features. In this research, to estimate the yield of canola, a yield database along with the time series of the Landsat and MODIS data of Moghan Agro-Industry Company fields were provided. Then, 10 spectral indices were evaluated for estimating the canola yield. The relations between the canola yield and the candidate indices were investigated and it was revealed that, during the flowering period, the NDYI index obtained a higher accuracy compared with other indices (r = 0.73). The fusion of the Landsat and MODIS time series data based on Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM), resulted in a 7%-increase and an 11%-decrease in correlation and RMSE (kg/ha), respectively. This research indicated that data fusion techniques are able to improve the performance of spectral indices and hence increase the accuracy of crop yield estimation.