Land degradation is a worldwide problem with natural and anthropogenic causes. An important anthropogenic cause of land degradation is land uses that deviate from land capability, which is the “natural” use. Despite the conscience of scientists about this issue, studies are lacking that set up a quantitative nexus between land capability and related drivers. This nexus is essential because it allows anticipating changes in the drivers that promote or recede degradation. In this study, a partial least squares–path model (PLS–PM) was used to help closing this gap. The pilot study occurred in seven Paraopeba River sub‐basins located in the State of Minas Gerais (Brazil), where capability is dominated by mosaics of forests and pastures. Land capability was assessed by the ruggedness number, being inversely proportional to it. Among multiple potential drivers of land capability initially included in the model, collinearity and other consistency and performance analyses indicated the percentage of latosols and (re)forested areas, annual rainfall, and water‐related processes (e.g., weathering, nutrient leaching) as prominent in the studied region. The results linked increases of latosol and forest occupation to runoff reduction [runoff = latosol × (−0.640) + forest × (−0.156) + reforestation × (−0.379)], suggesting an attenuation of erosion by these parameters, namely gully erosion that impacts on the ruggedness number and land capability in the sequel. The weathering of carbonate rocks was also related to land capability changes because the PLS–PM model exposed a relationship between the ruggedness number and products of carbonate rock dissolution: ruggedness number = alkalinity × (−0.578) + total magnesium × (0.462) + dissolved oxygen × (−0.418). In addition to highlighting processes capable of changing land capability overtime, the model equations brought attention to measures that could improve it in the region thus preventing degradation. The measures included management practices capable to raise the soil's dissolved oxygen (e.g., through aeration) or preventing soil's magnesium deficiency (e.g., through foliar sprays). The network of cause‐and‐effect relationships set up by the PLS–PM model was finally used to elucidate about potential land use changes that would bring capability in the Paraopeba River basin towards a better status, such as conversions to pasture land. Future work is expected to further elucidate about the benefits for land capability of specific land use changes, through the testing of expected socioeconomic development scenarios.