A Research on Cohesion Hyperspectral Detection Model of Fine-Grained Sediments in Beichuan Debris Flow, Sichuan Province, China

Cohesion is the main inter-controlled factor for the stability of fine-grained sediments in debris flow, and plays an important role in debris flow hazard early warning. At present, there is no cohesion rapid remote sensing detection model, which seriously affects the development of quantitative evaluation on debris flow stability. How to use remote sensing to quickly detect the cohesion of fine-grained debris has become an important scientific issue. Therefore, strengthening the research on the cohesion hyperspectral detection model, indicating its sensitive spectral bands, and establishing a quantitative model between cohesion and these bands are of great significance not only in discovering the stability mechanism, but also in quickly establishing the stability detection model for gully sediments. Taking the Beichuan debris flow as the study area, we carried out experiments on cohesion, cohesion influencing factors, and spectra. Firstly, six cohesion hyperspectral sensitive bands are indicated in red, near infrared portions of the electromagnetic spectrum, including 750, 1578, 1835, 2301, 2305, and 2309 nm; secondly, these bands discover the cohesion influencing factors. Band 750 nm indicates the characteristics of cohesion, effective internal friction angle, and permeability coefficient, while the other five bands indicate the characteristics of effective internal friction angle, density, and moisture; finally, a hyperspectral remote sensing detection model for the fine-grained sediments cohesion is established. With a correlation coefficient of 0.56, and p value less than 0.001, the model indicates that cohesion has a great significant correlation with the six bands. This not only provides sensitive bands for detecting cohesion of fine-grained sediments using remote sensing, but also provides a scientific basis for rapid detection of the fine-grained sediments' stability in large areas.