Abstract: Soil spectra under different intensities of human disturbance were used as the data sources to extract the absorption characteristics and sensitive bands by the continuum removal method. Linear regressions were established for estimating soil salinity based on the continuum removal method. The results showed that: ① The continuum removal method could be used to significantly enhance the absorption characteristics of soil spectra in wavelength bands of 400-610 nm and 610-800 nm; ② The correlation between soil salt content in the manned/unmanned disturbance area and the first derivative of the continuum removal (FDCR) was better than continuum removal (CR), the sensitive bands were concentrated in the Vis area in the unattended area (400-800 nm), and the artificially disturbed area was Vis-NIR (400-2 400 nm); ③ Basically, the accuracy of the models for predicting soil salinity in the manned/unmanned disturbance area was in an order of 6 absorption characteristic models of continuum removal>sensitive band model of CR and FDCR>3 absorption characteristic models of continuum removal>single absorption feature model. The values of R² of the best predictive model of the manned/unmanned disturbance regions were 0.86 and 0.92, respectively. The value of the optimal model of relative percent deviation for the unmanned disturbed zone reached 3.06 and higher than 3, which revealed that the model was excellent and stable in predicting soil salinity in this area.

Key words: human disturbance, spectral continuum removal, soil salinity, hyperspectrum, Xinjiang