The importance of soil structure data for soil erosion modelling and mapping

Soil structure is a crucial parameter for comprehending soil functioning, as it influences the impact of soil erosion and the behaviour of soil hydraulic properties. Simultaneously, soil structure is a crucial indicator of potential soil degradation or soil health status. Various proxies, including soil organic matter and soil texture, are employed to characterize soil structure in modeling. Nevertheless, the effectiveness of these proxies in determining soil erosion or soil hydraulic properties remains uncertain. To address this gap, we leverage available soil structure data along with diverse soil properties for Switzerland to investigate the relationship between proxies, soil structure data, and soil erosion rates. To achieve this, we initially developed a soil structure class map using approximately 42,000 samples and a random forest machine learning algorithm at various depths. The confusion matrices revealed accuracies ranging from 57% to 71% for different depths. Subsequently, we incorporated these soil structure maps into the soil erosion model and compared the results with those obtained using proxies. We anticipate that utilizing real measurements as opposed to proxies may lead to significant differences in erosion rates.