Multiple soil properties have a correlation with its water content (WC), thus determining soil moisture is a measure that allows the implementation of activities for proper crop administration. Therefore, a fast and low-cost technique is necessary, which allows efficient soil monitoring as a tool to make relevant decisions about its use and management. The objective of this study was to generate a spectral prediction model of three soils of different order (an Oxisol from the Meta department, an Inceptisol from the department of Cundinamarca, and an Alfisol from the department of Tolima), under WC of 0%, 15% and 30%. All samples were taken and carried to desired WC and then, the spectral curves were obtained by means of diffuse reflectance spectroscopy. We were obtained one model for each soil order, in all cases, the models show a good predictive capacity (R2 greater than 0.85 and RMSE less than 0.04). The principal components analysis showed that there can be a remarkable difference between WC for each of the soils.