Abstract: Abstract: Crops biomass has been one of the most important indicators to predict the plant growth status and crop yield. This study aims to estimate the dry biomass of the winter wheat aboveground using optical satellite remote sensing. The winter wheat biomass was acquired at four growth stages (jointing, heading, flowering, and filling stage), and three nitrogen treatments (N1, N2, and N3, application rates of 0, 150 and 300 kg/hm2) in 2011, 2012, and 2014 at the agricultural meteorological experiment station, Nanjing University of Information Science and Technology, Nanjing, China. Simultaneously, the narrow-band spectral reflectance of canopy was also collected from the winter wheat using Analytical Spectral Device (ASD). Afterwards, the hyperspectral remote sensing data was resampled into the broad band reflectance of RapidEye, Sentinel-2, and WorldView-2 satellites with the red edge bands using the satellite spectral response functions. All possible band combinations of Normalized Difference Vegetation Index like (NDVI-like) were validated in the different growth stages and nitrogen treatments. Meanwhile, the satellite broad-band reflectance was used as the inputs for the six machine and deep learning for the biomass estimation, including the Random Forest (RF), Support Vector Regression (SVR), Partial Least Squares Regression (PLSR), Deep Neural Network (DNN), Long- and Short-Term Memory recursive neural network (LSTM), and one-dimensional Convolutional Neural Network (1D-CNN). Finally, the models were developed using Leave-One-Out cross validation under different growth stages and nitrogen treatments. The results showed that the optimal NDVI-like vegetation indices that derived from the two arbitrary bands presented the highest sensitivity to the winter wheat biomass at the flowering stage (coefficient of determination, R2=0.50-0.56). It was also difficult to accurately estimate the biomass of winter wheat using only one vegetation index in the whole growth period. The nitrogen treatments were dominated the correlation between the vegetation indices and winter wheat biomass. Specifically, the high nitrogen treatment was enhanced the sensitivity of vegetation index to the winter wheat biomass. The vegetation index with the Sentinel-2 bands performed better than that with the rest. The R2 was over 0.50 between the vegetation index and biomass at the jointing and flowering stages. The best performance was achieved in the estimation model of winter wheat biomass using DNN among the six models. The performance of DNN-based model was also depended on the growth stages and nitrogen treatments. In the single growth stage, the highest estimation accuracy was observed at the jointing stage (R2=0.69-0.78 and the normalized root mean square error (NRMSE)=0.26-0.31), and at the flowering stage (R2=0.69-0.70 and NRMSE =0.24-0.35). The highest estimation accuracy was obtained in the DNN-based model with the Sentinel-2 bands as the inputs, indicating the R2 of 0.70 in the whole growth period. The high nitrogen treatment was also enhanced the estimation accuracy of DNN model, where the R2 was not lower than 0.71 and RMSE was within 219 g/m2 at the N3 condition for all the three satellite bands. Therefore, the optical satellite remote sensing data can be expected to estimate the winter wheat biomass under the different growth stages and nitrogen treatment conditions.