Abstract:Optimally using satellite carrying channels of blue (B), green (G), red (R) and near-infrared (NIR) to estimate crop N status play a crucial role in the management and estimation of regional N cycling. The current study was aimed to assess the performance of optimized normalized and integrated spectral indices, derived from simulated broadband GF-2 (GF) and Sentinel 2A (S) satellites data, to remotely sense plant N concentration in potato (Solanum tuberosum L.). Different field experiments were conducted with different N levels for two potato cultivars in Wuchuan County and Siziwangqi County at the northern Yinshan in Inner Mongolia from 2014 to 2016. The canopy reflectance data of potato at the growth stages of tuber formation, tuber bulking and starch accumulation were collected by a canopy hyper-spectrometer tec5. The estimation models of potato plant nitrogen concentration based on different spectral indices were constructed and validated by independent field data. The results showed that the extractive sensitive bands were able to guide the selection of satellite carrying channels. The optimal multi-spectral indices GF-GBNDSI and S-GBNDSI calculated based on the green (G) and blue (B) channels selected from the sensitive central bands had the highest coefficient of determination (R2) with plant N concentration of potato, and the R2 was 0.41 and 0.38, respectively. The multi-spectral indices GBNDSI/NDVI and GBNDSI/GNDVI constructed by combining NDVI and GNDVI with GBNDSI, respectively, could significantly improve the explanation ability of plant N concentration of potato. The R2 of GBNDSI/NDVI and GBNDSI/GNDVI based on GF-2 and Sentinel 2A channel ranged from 0.54 to 0.57. Compared with the red edge multi-spectral index, GBNDSI/NDVI and GBNDSI/GNDVI not only overcame the lacking the red edge channel with most high-resolution satellites, but also reached a better estimating ability like red edge based REBNDSI/NDVI (R2=0.53) and REBNDSI/GNDVI (R2=0.59). The validated results showed that the root mean square error and mean relative error of the S-GBNDSI/NDVI and GF-GBNDSI/NDVI models were about 0.40% and 10.27%, respectively. Since most high-resolution satellites, especially most of the domestic satellites lacking the red edge channel, the optimized GBNDSI/NDVI and GBNDSI/GNDVI involving conventional channels can be used to monitor plant N concentration in crop.