The first modern soil science studies involving the measurement of soil properties such as those made by Professor Von Liebig, focused on the quantification of pure mineral elements present in the soil matrix. As the first half of the 20th century passed, the first works in the area of soil spectroscopy took place and the number of measurable soil attributes went from less than ten to many thousands. As an inmediate consequence, the use of multivariate methodologies and computer-based analysis became indispensable. Today, we are witnessing a moment in history where the amount of soil information available probably exceeds our analytic capabilities; therefore the need for transforming this raw data into useful information for final users is a continuous challenge. This thesis embodies a collection of novel studies related to the use of multivariate information provided by spectroscopic tools such as Visible and Near Infrared (Vis-NIR) spectrometers to represent soil variation. The thesis general structure is organized to follow the increasing levels of soil complexity, starting from the characterization of soil aggregates and the identification of soil colloids, to the recognition of soil horizons and their boundaries in the soil profile, to finally the depiction of soil type’s distribution in the landscape. It is shown, as the complexity of the soil target attribute increases (soil colloids < soil aggregates < soil horizons < soil types), the use of the multivariate information needs to be adapted as well, starting from recognition of individual soil properties (identification of individual soil colloids represented in a single spectrum in Chapter 2) to the use of the full variation contained in the spectroscopic information (creation of multidimensional indices in Chapter 5). Briefly, Chapter 1 is written as a rationale, presenting the need for up-to-date methodologies for making effective use of the increasing amount of soil information produced worldwide. It gives examples of the existing types of soil information and suggests the use of soil spectroscopy and multivariate analyses techniques as candidates for the study of key aspects of soil at different scales and levels of complexity. Accordingly, starting from the micro-scale and the analysis of individual soil properties, Chapter 2 presents a newly developed methodology for the measure of soil aggregate stability and the further use of spectroscopic information to predict its values. The new methodology makes use of an image recognition algorithm that allows measurement of the slaking of soil aggregates in water over time. The resulting increase of area in time was fitted to a function and three “slaking” coefficients were obtained. Coefficient a, was related to the highest possible area increase due to slaking in time. Coefficient b was associated with the initial time of slaking and finally coefficient c was linked to the rate of area increase in time. It was found that these coefficients were related with known soil properties, for example, coefficient a was positively related with exchangeable sodium (cmol(+) kg-1) and negatively related to exchangeable calcium–exchangeable magnesium ratio (Ca Mg-1). On the other hand, coefficients b and c were related with percentage of total carbon (%TC) and percentage of total nitrogen (%N). The second part of this chapter presents the use of different types of spectroscopic instruments, namely Vis-NIR and Middle Infrared (MIR) both portable and bench spectrometers, for the prediction of the slaking coefficient a. The regions of the electromagnetic spectra identified as important, were associated with Fe oxides and Fe bearing minerals, organic carbon and presence of kaolinite bearing materials. This chapter also presents the possibility of predicting slaking values using Vis-NIR and MIR information, showing that if an appropriate calibration dataset is used, the predictions are moderately acceptable (R2 = 0.6). Following the same line of research, in terms of the prediction of individual soil properties by using Vis-NIR spectroscopy, Chapter 3 demonstrates the use of selected state-of-the-art spectroscopic modelling techniques and large spectral libraries. The example presented in this chapter, involves the use of two large spectral libraries containing soil samples from the conterminous United States for the prediction of %TC in samples of a certain region of U.S. or “local samples”. The results of this chapter put special emphasis on the large amount of variability that Vis-NIR information could contain and on the importance of a properly selected set of soil samples to create Vis-NIR models. Even though all the modelling approaches were successful (R2 > 0.9), the important effect of similar composition regarding their geographic similarity which significantly affected the final bias of the models was noticeable (0.8 vs 0.02 %TC). As the complexity of the target soil feature increases (soil colloids < soil aggregates < soil horizons), it is observed that rather than using multivariate information to predict separate soil properties, this information can be directly used to describe soil variation. In this sense Chapter 4 details the development of a new method for the identification of soil horizons and their boundaries using fuzzy clustering of Vis-NIR spectra. The results of this chapter showed that the newly identified horizons (spectrally derived horizons), bear a resemblance to those described by pedologists. Chapter 5 follows the same essential concepts of Chapter 4 in terms of using the raw information contained in the Vis-NIR region of the spectra to characterize the soil profile rather than using individual soil attributes. This chapter expands into a new way of measuring the diversity of soils in the landscape. It presents two new indices for measuring soil functional diversity or “Functional Pedodiversity” inspired by previous studies in Functional Ecology where individuals are characterized in a multidimensional space composed by several continuous properties. The new “Functional pedodiversity indices” are then calculated using information from the whole Vis-NIR spectra and then are compared with the conventional approaches or “Taxonomic pedodiversity indices” which use previously classified soil taxonomic orders. The results of this chapter show the close connection between soil attribute variation and soil taxonomic units. This chapter offers a new perspective on the measurement of pedodiversity and represents the first study that assesses pedodiversity by spectral means. Finally Chapter 6 discusses the main findings of this thesis and foresees issues, challenges and opportunities in the area of spectroscopy and multivariate soil data analysis.Thanks to my parents Mario and Taly and my sisters Ivonne and Pauli, I love you all, and I’ll be there soon to say LLEGUÉ! It’s not just about doing a PhD, it’s about to be able to do it!