Soybean canopy nitrogen monitoring and prediction using ground based multispectral remote sensors

Remote sensing techniques applied in crop monitoring and management can help to reduce the input of nitrogen without reducing crop yield and accurately predict nitrogen demand [1]. The objective of this study is to use the ground based multispectral images to predict canopy nitrogen level for soybeans in southwestern Ontario. A light weight multispectral camera were used to collect multispectral measurements for four soybean fields from July to September in 2015. An evaluation of existing nitrogen indices were carried on in this study for soybean canopy nitrogen to select the best fit index for the study area. The results show that the modified RENDVI ^780-730 has the best correction between soybean nitrogen level and the spectral based index, the R ² is 0.70. This index is sensitive to vegetation structures Leaf Area Index (LAI) which is a confounding factor for the remote estimation of nitrogen. This index will lead an inaccuracy nitrogen prediction for soybeans. Therefore, multi-linear regression (MLR) analysis method using five band information was carried on and established a canopy nitrogen model for soybeans in this study. The R ² of the model is 0.745 and the RMSE is 0.51.