Remote sensing estimation of winter wheat leaf nitrogen content based on GF-1 satellite data

Abstract: Nitrogen is a major element for plant growth and yield formation in agronomic crops. Crop nitrogen content estimation by remote sensing technique has been being a topic research in remote sensing monitoring of agricultural parameters. Hyper-spectral remote sensing with wealth of spectral information has been widely used in crop physiological and biochemical information extraction. It provides theoretical basis for estimating crop biochemical parameters based on multi-spectral satellite data. In terms of multi-spectral satellite remote sensing, spectral reflectances and spectral indices are effective ways to establish estimation models of biochemical parameters, but which bands and spectral indices are more effective and reliable for leaf nitrogen concentration monitoring in winter wheat is still debatable. In this article, ground-based canopy spectral reflectance and leaf nitrogen content (LNC) of winter wheat were measured from field and plot experiments including varied nitrogen fertilization levels and winter wheat varieties across the whole growth stages. Multi-spectral broadband reflectance was simulated by using the measured hyper-spectral reflectance and spectral response functions of multi-spectral camera of GF-1 satellite with a spatial resolution of 8 m, and then, they were used for the establishment of spectral index (SI). Eight spectral indices significantly correlated with LNC at the 0.01 probability level were used to construct the LNC estimation models in a linear, quadratic polynomial and exponential regression model respectively. Considering the influence factors in evaluating the efficiency of the SI–LNC model, i.e., the stability of the SI to other perturbing factors, the sensitivity of the SI to a unit change of LNC, and the dynamic range of the SI, the improved sensitivity index was proposed based on the NE and TVI index models. The optimal LNC estimation model was given according to the sensitivity and accuracy analysis, and the model was used to inverse the LNC in greenup growth period based on the GF-1 satellite image. The results showed that: 1) The simulated multi-spectral reflectance was highly correlated with the spectral reflectance from remote sensing images in visible and near infrared bands. They were consistent with each other keeping a correlation coefficient of greater than 0.95. It was concluded that the simulated broadband SI considering the spectral response function could be used to analyze the quantitative relationship with leaf nitrogen in both different growth periods and whole growth stage. 2) The SI based on the simulated spectral reflectance was significantly related with the LNC at 0.01 probability level with the correlation coefficient of greater than 0.6. A different pattern of the best combinations was found for 6 two-band spectral indices. The selection of 610-690 nm paired with 750-900 nm was the most effective two-band combination in RVI index, which was also the center wavelengths of the red and near infrared bands for GF-1 satellite data. 3) The sensitivity analysis indicated that all the regression models of selected SI passed the significance test at 0.01 probability level. The TCARI/OSAVI and RVI indices linearly related with LNC implied a stable response to the LNC changes. The first-order differentials of RVI and TCARI/OSAVI with respect to LNC were 9.44 and 3.08, and the sensitivity indices were 0.0671 and 0.1979 respectively. The RVI index was regarded as the most suitable index for LNC estimation. 4) The TCARI/OSAVI and RVI indices performed well in accuracy test, and the RVI index was more excellent in remote sensing mapping based on the GF-1 satellite image. Taking all factors into consideration, we believed the model based on the RVI index was optimal for LNC estimation with the determination coefficient of 0.6.