A fundamental aspect of precision agriculture or site-specific crop management is the ability to recognize and address local changes in the crop production environment (e.g. soil) within the boundaries of a traditional management unit. However, the status quo approach to define local fertilizer need relies on systematic soil sampling followed by time and labour-intensive laboratory analysis. Proximal soil sensing offers numerous advantages over conventional soil characterization and has shown potential for management zone (MZ) delineation and site-specific crop management. While electromagnetic induction (EM) based sensors have been widely used, the use of spectroscopy-based sensors is still in its infancy. This study evaluated the capability of spectroscopy-based Veris P4000 soil sensor in predicting soil properties and two proximal soil sensors (EM-based DUALEM21S and Veris P4000) in delineating MZs from two commercial potato (Solanum tuberosum L.) fields from New Brunswick, Canada. The proximal sensor collected data (apparent electrical conductivity ECa and spectra) were then used to delineate MZs using the clustering method. The efficiency of the MZ delineation was then compared with the laboratory measured soil properties and the yield monitor data collected over three years. In total, 295 soil samples were collected and analyzed using standard laboratory procedures. DUALEM21S was used to map the ECa at four depths. Visible and near infrared (Vis-NIR) spectra (397-2212 nm wavelength) were collected in triplicate for all samples using the spectrometer from Veris P4000 system in laboratory conditions. The dataset was separated into calibration (70%) and validation subset (30%) and partial least square regression (PLSR) models with bootstrapping were developed and validated against laboratory measured soil properties. Spectroscopic system well predicted soil properties. It showed the strongest potential to predict soil organic matter (SOM) with the highest accuracy. Other soil properties that were either positively (e.g. Ca, Mg) or negatively correlated (e.g. pH, buffer pH) with SOM were also predicted with good accuracy. The point samples were then interpolated to field boundaries and used to develop MZs. The number of MZs were then optimized following the normalized classification entropy (NCE) and the fuzziness performance index (FPI). While lab measured physio-chemical properties identified three optimum MZs, the spectra predicted properties and DUALEM21S data identified two MZs. The choice of two MZs was consistent with the number of MZs identified from the yield monitor data. Clay content, soil moisture content and P concentration showed strongest correlations with yield variability among soil properties. DUALEM21S data from all four depths were strongly correlated with yield. In contrast, measured spectra showed relatively weak prediction (except principle component 1, PC1 and PC3) of yield. These results suggest that the EM based sensor was effective in delineating MZ, whereas more advancements are required to use spectra in MZ delineation.