The objective was to identify landscape areas with different patterns of variability using a statistic protocol with data of magnetic susceptibility (MS) and soil color that are covariate attributes of soil formation factors and processes. The studied area, of 380 ha, is located in Northeast of São Paulo State, Brazil. An amount of 86 samples was collected using 30 m intervals on the transect. At the transect sides, 150 samples were collected at 159 m intervals (a point each 2.5 ha). First the accuracy limits have been validated in the transect using the technique of Split Moving Windows — SMW. The limits identified in the transect were extrapolated to the sides using the contours of variability maps. The MS peaks SMW, for both depths, presented a correlation with the peaks of clay content (r = 0.7; P < 0.01), hue (varying from 0:37; P < 0.05 to 0.61; P < 0.01) and Normalized Difference Vegetation Index—NDVI (varying from − 0.25 to − 0.35, P < 0.05). The errors of the MS spatial variability maps (6.22–11.85%) were similar to the clay content ones (6:22 to 14:16%). MS was more efficient in the compartmentalization of the landscape (identification of areas with different patterns of variability) than the hue determined by diffuse reflectance spectroscopy in Oxisols under the transition Basalt and Colluvial–Elluvial–Alluvial Deposits. The results of this study can lead to using an alternative strategy that is a mapping of soil attributes and identification of areas with different patterns of pedogenic iron oxide variability.