Sequential indicator conditional simulation and indicator kriging applied to discrimination of dolomitization in GER 63-channel imaging spectrometer data

Abstract

Laboratory reflectance spectra of synthetic mixtures of the carbonate minerals calcite and dolomite were measured in the visible and near-infrared wavelength region (0.4–2.5 μm) using a high-spectral resolution laboratory spectrometer. The instrument measured reflectivity with an accuracy of 0.001 μm, allowing detailed resolution of the carbonate spectrum. The spectra of calcite and dolomite could be characterized by the presence of a strong absorption band centered at 2.3465 μm for pure calcite and at 2.3039 μm for pure dolomite. Nine mixtures of intermediate composition were analyzed demonstrating that the position of the carbonate absorption band is semilinearly related to the calcite content of the sample. Theoretically, this model allows mapping of dolomitization from high-spectral resolution remotely sensed imagery, GER 63-channel imaging spectrometer data from southern Spain were used to attempt such a mapping. First, pixels of vegetation were removed. For the remaining pixels, the wavelength center of the carbonate absorption band was detected and converted to a category of calcite fraction. The percentage of calcite for the remaining pixels was estimated using direct indicator kriging (IK) and sequential conditional indicator simulation, assuming that the calcite content could be represented as a category variable (SCIS category variable) and as a continuous variable (SCIS continuous variable). Four realizations of an SCIS (category variable) showed that on the average, 60 percent of the data was simulated in the same class and over 90 percent of the data within one class difference. A comparison with field samples showed that IK estimates of calcite content were within 20 percent accurate. The SCIS (continuous variable) does not perform as well with differences between −45% and +26% calcite; however, simulation reproduces the spatial variability better.