Tracking the source of eroded sediments causing siltation in the Ibirité Lake, Brazil, via proximal sensing and soil variability

Soil erosion is a global concern given its negative consequences to the quality of soils and water bodies. This issue can be mitigated through proper soil management and identification of the sources of sediment causing erosion and siltation. This work aimed i) to characterize soils of an urban/rural watershed in Minas Gerais, Brazil, and ii) to track the source of sediments reaching the lake present in this watershed. Samples of soils (86) and sediments (16) were collected throughout the watershed (91.6 km²) (0-2 cm of depth) and at the bottom of the Ibirité lake (2.8 km²), respectively. About 40% of the watershed is occupied by urban areas, with the remaining land uses/land cover being divided by native vegetation, pasture, and crops. Soils were classified in every sampling place after morphological description. Texture (clay, silt, and sand contents) of soils and sediments were determined by the pipette method along with their total elemental contents, through a portable X-ray fluorescence spectrometer (Tracer 5g, Bruker), and their magnetic properties, using a Bartington MS2B magnetometer. Then, groups of sediments were separated by cluster analysis based on their texture, chemical composition and magnetic properties. These clusters were validated by comparing the samples within each group with their morphological properties. Soil properties were spatialized to the entire watershed using the Multilevel B-splines method (SAGA GIS software) and sediment clusters were applied to these soil property maps, identifying the similarities between sediments and soils across the entire watershed. Finally, this map was combined with areas containing bare soil and topographic features that determine the hydrological connectivity of streams from those places to the Ibirité Lake, thus creating a map showing the areas more prone to be the source of sediments. Fe, Si, magnetic properties and texture were the most important properties to differentiate clusters of soils and sediments. Clusters using only texture were less satisfactory to separate those samples. Groups of soils with similar physical and chemical properties could be separated, indicating their parent materials. These physical and chemical properties are related to the mineralogy of these samples and, hence, could be used as fingerprinting of the soils being eroded. Topographic features and bare soil areas along with soil physical and chemical properties indicated the places more likely to produce sediments. Although there is a great impact of urban areas on the production of sediments, which can change over time, the application of these sediment clusters to the soil property maps for the entire watershed determined the most likely source of sediments causing siltation, confirming the potential of this proximal soil sensing approach to this end. Future works on such large and constantly changing areas should focus on time-series analysis and the search for other environmental variables to help track the source of sediments in such close-to-urban areas.