Quantitative analysis of convolved thematic mapper spectra of soils in the visible near-infrared and shortwave-infrared spectral regions ((m-l’sum)

Laboratory reflectance spectra of 91 soil samples from Israel were convolved into the six Thematic Mapper (TM) bandpasses allocated within the VIS-N1R-SWIR spectral region (0-4-2-5/µm). The TM spectra were used as raw data against which to run the NIRA-VNIRA (near-infrared analysis and visible and near-infrared analysis, respectively) procedure using a method called Thematic Mapper Analysis (TMA). Several soil properties and constituents were examined for possible prediction from the TM spectra, but only four were found to be significantly predictable: carbonates (CaCO₃), specific surface area (SSA), total silica (SiO₂) and loss-on-ignition residual (LOI). The prediction of each property was based on the total apparent reflectance value (albedo) rather than specific absorption. The intercorrelation between the properties (particularly with CaCO₃) is considered the major pathway in which the spectral-based prediction took place. Prediction performance of TMA is still low relative to the results obtained using a similar population and an analysis routine for highest spectral resolution data. Under the extremely difficult conditions of the TMA run (low spectral resolution and complex soil matrix) the prediction performance is likely to be good and is very promising for the remote sensing applications of soils. Further study with real TM data and field measurements is strongly recommended.