This study aimed to derive mathematical models to predict the soil organic matter content based on soil color obtained by a colorimeter in the Munsell color system. A total of 907 soil samples were collected in the region of Porto Grande (Amapá, Brazil) and analyzed in the laboratory for chemical properties, particle size distribution and color of dry and wet samples. The Munsell color components value and croma obtained using a colorimeter were used to predict soil organic matter content based on stepwise multiple linear regression. Models derived using all samples had R² of 0.66 for wet samples and 0.56 for dry samples, respectively, when validated using independent samples. It was possible to improve the models by separating the samples by soil class or texture. The models derived using colors obtained from wet samples were systematically better than those based on dry samples. Among soil classes, best results were obtained for Argissolos (Ultisols) and Latossolos (Oxisols), both having an R² of independent validation of 0.73 (wet sample). For texture, best results were obtained for very clayey soils, with an R² of validation of 0.81 (wet sample). The soil organic matter prediction models based on soil color have simplicity and potential to be used in the laboratory and in the field with quick and unnecessary chemical products, especially for Ultisols and Oxisols of clayey texture.
texture; soil class; soil color; colorimeter
