An Advanced Analytical Approach for Spectral-Based Modelling of Soil Properties

Nimrod Carmon, Eyal Ben-Dor

Research output: Contribution to journalArticlepeer-review


developing accurate and robust prediction models to analyse soil attributes from spectral information has a significant importance for hyperspectral remote sensing applications. Using partial least squares for model development is a multistep process with many optional alterations. Although crucial for the result, pre-processing algorithms to be applied to a given dataset are usually selected in a non-systematic procedure that ends once the user obtains a favourable result based on a subjective impression. These results are sensitive to many aspects of model development, including grouping method, validation technique, pre-processing calculation and model statistical parameters, among others. In this study, we developed an optimal and automatic systematic procedure for model development that takes into account many possible alternatives, and includes a novel pre-processing technique and model-validation approach. Based on the many options available to extract a suitable model, we developed an automatic data-mining machine and parameter set to judge the results and the physical assignments used by the model, in order to extract the best model for practical remote sensing applications. An evaluation tool for correlations between spectral and modelled data is demonstrated to highlight the power of the suggested approach. The developed system, termed PARACUDA II® was tested on the legacy soil spectral library of Ben-Dor and Banin, which had been used to establish the soil chemometrics approach.
Original languageEnglish
Pages (from-to)90-97
Number of pages8
JournalInternational Journal of Emerging Technology and Advanced Engineering
Issue number3
StatePublished - 3 Mar 2017


Dive into the research topics of 'An Advanced Analytical Approach for Spectral-Based Modelling of Soil Properties'. Together they form a unique fingerprint.

Cite this