Reflectance spectroscopy is an alternative to describe soil properties, with potential to reduce costs and environmental impacts of conventional practices related to this activity. Acquisition of soil spectra on-the-go has several advantages over 'in-situ' static approaches, like deriving information with high spatial density. However, issues concerning on-the-go spectral measurements exist, mainly due to sensor movement and heterogeneous soil condition in the field. Procedures to mitigate these drawbacks, like external parameter orthogonalization (EPO) and direct standardization (DS), have mainly been applied so far to static spectral readings. In this study, EPO, DS and orthogonal signal correction (OSC) are tested in the context of on-the-go spectra acquisition for prediction of soil properties related to liming (i.e., pH in CaCl2, pH in SMP buffer, concentration of organic matter, calcium and magnesium, potential acidity, sum of basis, cation exchange capacity and its saturation by basis, lime requirement and moisture content). A detailed dataset (300 soil samples coupled with laboratory and field spectral measurements) was acquired in two sites in Brazil with contrasting soil attributes (site 1 with ‘clayey texture’ – Ferralsol; site 2 with 'sandy texture’ – Alisol), and variability of soil properties was increased in these sites through application of different limestone rates in experimental plots. Spectral correction procedures slightly improved the accuracy of lime requirement predictions, with reduction of root mean squared error (RMSE) from 1.43 to 1.17 t ha−1, for study site 1 after applying OSC, and from 0.59 to 0.44 t ha−1, for study site 2 after DS was implemented. However, models based on laboratory data still performed considerably better with RMSE of 0.99 and 0.43 t ha−1 for site 1 and 2, respectively. ‘Global’ (i.e., one general correction model for a given field) or 'specific’ models (i.e., several correction models, derived according to clusters obtained through fuzzy k-means applied to OSC components) performed considerably worse in comparison with other studies. Probably occurrence of external factors affecting the spectral information was not constant in the mapped fields. Also, different external factors may have affected the spectra at the same time and efficiency of the correction procedures decreased. Considering the high sensitivity of predictions based on field data to the approach used to interpolate the spectra and the poor performance of the correction methods applied in this context, more investigation is needed to improve predictions based on spectral data acquired on-the-go. Homogeneous spatial distribution of factors not related to the properties of interest, or at least in a degree allowing correction by the methods tested here, may not happen when current measurement systems are used. Despite that, spatial patterns described by wet-chemical analysis could be represented, at certain extent, through predicted values of lime requirement (LR), derived from field spectra. For instance, predictions after spectral correction resulted in autocorrelation patterns and map of LR comparable with those observed using conventional methods, for the site 1. These results, coupled with a semiquantitative potential of predictions based on field data after spectral correction, indicate that on-the-go measurements have potential for soil properties characterization, although full quantitative potential will require further advances in sensing solutions and chemometric methods applied in this context.