Detection of disturbed forest ecosystems in the forest-steppe zone using reflectance values

Abstract

This paper presents results of the assessment of discriminant analysis potentialities for detecting disturbed forest ecosystems in the forest-steppe zone using their reflectance spectrum properties. A new method is proposed for the automated detection of disturbed forest stands among forest-covered lands, based on the discriminant analysis of the magnitude of changes in the reflectance in various spectral ranges. Using experimental data from 1836 forest areas typical of the forest-steppe zone of the Central Chernozem region, we propose equations that allow a specific forest area to be classified as disturbed or undisturbed forests in an automated mode. The accuracy of disturbed forest detection is about 90%. It is found that variations in the short-wave infrared reflectance are most informative for disturbed forest land detection when compared with the reflectance variations detected by the Landsat sensors in the other spectral ranges.

Keywords

• disturbed forest ecosystems
• stepwise discriminant analysis
• remote sensing
• Landsat
• reflectance spectrum properties