Scholarly record

ANALYSIS OF BIG DATA FOR ESTIMATING THE INFORMATIVENESS OF THE COEFFICIENTS OF THE MULTI-TEMPORAL SOIL LINE N-DIMENSIONAL SPACE

Alexey Rukhovich

Abstract

An algorithm for extracting information on the types and subtypes of soils for each pixel of big data sets of remote sensing on agricultural lands is proposed. For this, a mathematical apparatus for constructing a multi-temporal (empirical) soil line is used. The classical soil line (SL) is defined in the RED-NIR spectral space by two coefficients "a" and "b" [2, 6, 9]. In this form, SL does not characterize soil types and subtypes. A multi-temporal soil line (MSL) is the major axis of an ellipse that describes all possible pairs of RED-NIR values that a pixel (resolution element) of remote sensing data corresponding to bare soil surface can take. MSL is specified in the spectral space by several (number N) coefficients. The resulting N-dimensional space of MSL coefficients makes it possible to give a unique characterization of each type and subtype of soils in the following zonal series: soddy podzolic, light gray forest, gray forest, dark gray forest, podzolized chernozems, leached chernozems. Dispersion analysis allows us to state that the soils of this series are statistically separated from each other in a set of MSL coefficients. That is, the MSL coefficients are characteristic of the type and subtype of the soils, and MSL is an empirical soil line (ESL) of the soil type and subtype. Classical SL is a collection of empirical soil lines of different soils on one remote sensing data scene.

Publication details

Back to publication list