This study focuses on soil total nitrogen in Fuzhou City.Model input variables were firstly se-lected and optimized on the basis of soil reflectance and 13 transformed spectra.Then 20 combined spectral input models were constructed to invert soil total nitrogen in the study area.The results show that 20 models using combined spectral input variables have high prediction accuracy and stability.A-mong them,models that can accurately estimate total nitrogen content account for about 60.0%and those that can roughly estimate the content 40.0%,an increase of 59.3%and 6.4%respectively com-pared with single spectral models.It indicates that soil total nitrogen inversion models constructed with combined spectra enable different transformed spectra to complement each other,enhancing both pre-dictive capability and model stability.This way of modeling gives new insights into hyperspectral predic-tion of soil total nitrogen.
维普
万方数据