Developing and applying novel spectral feature parameters for classifying soil salt types in arid land. (July 2015)

• Record Type: Journal Article
• Authors: Jin, Pingbin; Li, Pingheng; Wang, Quan; Pu, Zhi
• Is Part Of: Ecological indicators. Volume 54(2015)
• Page Start: 116
• Page End: 123
• Publication Date: 2015-07
• Subjects: Arid land -- Soil salinity -- Salt type -- Hyperspectral -- Feature parameter -- Classification algorithm; Environmental monitoring -- Periodicals; Environmental management -- Periodicals; Environmental impact analysis -- Periodicals; Environmental risk assessment -- Periodicals; Sustainable development -- Periodicals; 333.71405
• DOI: 10.1016/j.ecolind.2015.02.028
• Languages: English
• ISSNs: 1470-160X

Abstract

Abstract: Soil salinization is a major desertification process that threatens especially the stability of arid ecosystems. There is an urgent need for intensive monitoring and quick assessment of salinization through remote sensing as a tool for combating desertification in such ecosystems. Recent researches have revealed that in order to retrieve soil salt contents accurately from hyperspectral reflectance, a pre-knowledge of salt types is required, which greatly outlines the spectral features of saline soil reflectance. In this study, a set of feature parameters have been developed after a thorough investigation of spectral responses to different soil salt types and salt contents for quick and accurate classification of soil salt types. The application has been validated using three independent datasets composed from: laboratory experiments (dataset I), in-situ field measurements (dataset II), and satellite-borne Hyperion image (dataset III). For comparison, four other common classification algorithms have been validated using the same datasets. The results showed that the new approach proposed in this study performed well with not only single-type but also multiple-type salts for which the four common algorithms performed rather fairly. Furthermore, validating using datasets II and III showed that the newly proposed approach had a stable performance while the other four failed, indicating the advantage of the new approach. The feature parameters developed in this study hence provide a novel and efficient approach for salt type classification from reflectance spectra, and we foresee its potential applications on large-scale soil salt type mapping towards better understanding soil salinity characterization from remote sensing data.