Brazil and the world need to use soils appropriately and there are no pedological or soil attributes maps on a semi-detailed and detailed scale suitable for proper management, according to necessity of agriculture. Traditional soil mappings are expensive and time-consuming. In order to map large areas in short time it is necessary to use technology, hence, remote sensing technique in orbital platforms are promising. Thus, it is expected that spectral characteristics of bare soil surface obtained from Landsat TM sensor have positive relation with soil texture. This work aims to generated a predicted model to quantify clay based on bare soil spectral reflectance obtained from Landsat TM image and granulometric soil analysis collect at the field. Pre-existing maps, such as the pedological map (SCALE 1:100.000) and geological map (SCALE 1:250.000) were used for validate the spectral data, as well as the samples collected in the field. The results show that the bare soil map obtained from Landsat were efficient and suitable for modeling soil surface clay. The soil surface clay model present good performance reaching R2 of 0.68 and RPD>2.0. Therefore, the clay map has shown to be appropriate to refine soil map survey and improve pedological and geological studies.