Exploring new spectral bands and vegetation indices for estimating nitrogen nutrition index of summer maize

摘要

Accurately and timely diagnosis of plant nitrogen (N) status is imperative for N fertilization management and yield prediction of summer maize. This study was aimed to identify the most sensitive/appropriate spectral band combinations to estimate the N nutrition index (NNI) by comprehensive analyses on canopy spectral reflectance from visible to near-infrared light, to develop the optimum vegetation indices for NNI during V6-V12 growth period, and to validate the regression models for estimating NNI of summer maize by comparing the two methods (direct and indirect) to determine the most appropriate method for practical use. Five multi-locational and multi-N rates (0-320 kg ha(-1)) field experiments were conducted during three growing seasons (2015, 2016 and 2017) using five summer maize cultivars. The measurements regarding canopy spectral reflectance, plant biomass, and plant N concentration were taken at critical stages of summer maize under the various N treatments. Comprehensive analyses on the different regression models of NNI for normalized difference spectral index (NDSI) and ratio spectral index (RSI) composed of any two bands between 325 and 905 nm of summer maize were made by using the reduced precise sampling method. The NNI values in the present study ranged from 0.68 to 1.15 under different N treatments. The most sensitive spectral bands were located at 710 nm (red edge band) and 512 nm (visible light band) and the optimum spectral vegetation index for estimating NNI was NDSI (R-710, R-512). The linear regression model between NDSI (R-710, R-512) and NNI was NNI = 0.95 NDSI (R-710, R-512) + 0.14. Additionally, the soil-adjusted vegetation index (SAVI) was used to correct NDSI(R-710, R-512), and the performance of the linear regression model was best when the parameter L (soil-brightness correction factor) of SAVI (R-710, R-512) was 0.05. The performances of the direct and indirect NNI estimation methods were compared. The validation results showed that the performance of the newly developed vegetation indices (NDSI (R-710, R-512) and SAVI (R-710, R-512)((L=0.05))) was the best with the relative root mean square error (RRMSE) values ranging from 11.4% and 13.1% in the direct method; while the performance of the existing vegetation indices (Ratio Vegetation Index II and modified SAVI) were best with RRMSE value of 16.9% in the indirect method. It was concluded that both the direct and indirect methods can be used to estimate NNI of summer maize, but the construction of the newly developed vegetation indices was easier in the direct method. The projected results will provide a technical basis for potential application of remote sensing technology for monitoring and diagnosis of plant N nutrition in summer maize production.