Application of mass balance and unmixing model to trace sediment sources in an agricultural catchment

Abstract

Sediment deposition gives rise to a variety of environmental issues, which pose a significant threat to ecological safety. Information regarding the sources of topsoil and sediment sinks in the catchment is crucial for understanding the soil erosion process and can aid in the development of targeted catchment management strategies. Fingerprinting models, such as the mass balance and unmixing models, are important quantitative tools for identifying erosion sources and sinks. This study aims to quantify sediment sources in an agricultural catchment and compare the effectiveness of the genetic algorithm-based mass balance model with the Monte Carlo-based unmixing model within the composite fingerprint technique for tracing sediment. Topsoil samples representing four land use types (i.e., slope farmlands, terrace fields, river banks, and road banks) and four geological units (i.e., J₂s, J₃sn, J₃p, and Qh) were selected as potential sediment sources, while sediment samples collected over three years from Changshou Lake were used as sink samples. The mass balance model results indicated that the average contributions from slope farmlands, terrace fields, river banks, and road banks were 70.8 %, 3.0 %, 22.1 %, and 4.1 %, respectively. Contributions from J₂s, J₃sn, J₃p, and Qh were 22.3 %, 66.6 %, 5.2 %, and 5.9 %, respectively. In contrast, the unmixing model results indicated that road banks and J₃sn were the primary contributors to sediment sources in the agricultural catchment. This can be attributed to factors such as steep topography, frequent agricultural activities, and road construction, which made slope farmlands and road banks the primary sources of sediment. Due to their intrinsic physical properties, J₃sn and J₂s weather more rapidly, have shorter migration paths, and are frequently subject to agricultural production activities, making them major contributors to sediment. Both models allow for the quantification of sediment sources through geochemical fingerprints. However, the genetic algorithm-based mass balance model was chosen for this study due to its higher goodness-of-fit (GOF) and lower uncertainties. The average GOFs of the mass balance model were 91.5 % for land use and 94.2 % for geological units. The findings of this study provide a scientific foundation for sediment source modeling and comprehensive catchment erosion management, as well as a more accurate guide for the application of the mass balance and unmixing models in sediment source studies.