Abstract: Abstract: Soil erosion has presented a major environmental threat to sustainable agricultural production in karst areas. Scientific control of soil erosion depends mainly on the sources of suspended sediment on the surface and suspended sediment in the underground rivers in the small watersheds in karst valleys. Among them, the typical karst areas are characterized by dual lithologic structures on the surface and underground. Taking the small watershed of Qingmuguan karst valley in Chongqing of southwest China as an example, accurate and rapid identification of the sediment sources was proposed for the targeted prevention and control measures of soil erosion. Soil samples were collected from four sediment sources, including carbonate rocks, clastic rocks, channels, and fissures. 41 soil physical and chemical properties were measured as the fingerprint factors, such as the radionuclide 137Cs, magnetic susceptibility (χ), Li, and As. The bilateral range and mean test were utilized to screen the 33 fingerprint factors. Further screening was performed on the 33 fingerprint factors to differentiate the sediment sources using the Kruskal-Wallis H nonparametric test. The fingerprint factors were screened with significant differences (P<0.05) among different sediment sources. A stepwise multivariate discriminant analysis was performed on the fingerprint factors with the significant differences between groups using the premise of Wilks' lambda minimization. The best combination of fingerprint factors was achieved in the Fe, 137Cs, Pb, V, and χfd. The composite fingerprint identification was finally used to quantitatively analyze the sources of suspended sediment in the surface and underground river outlets. The results show that there were certain regularities between rainfall, rainfall duration, rainfall intensity, and the contribution rate of potential sediment sources in the basin. The variation in the contribution rate of suspended sediment on the surface of carbonate rocks was proportional to the rainfall and rainfall duration. Specifically, the contribution rate of carbonate rocks increased gradually with the increase of rainfall and rainfall duration. There was no significant change in the contribution rate of suspended sediment on the surface of the channel with the rainfall. There was a gradual increase in the contribution rate of the carbonate underground suspended sediment, as the rainfall and rainfall duration increased. The contribution rate of fissures to the underground suspended sediment sources was inversely proportional to the rainfall, indicating a gradual decrease with the increase in rainfall. The average contribution rates of potential sediment sources in the watershed to the surface suspended sediment were ranked in the descending order of the carbonate rocks (43.62%) > channels (27.34%) > clastic rocks (17.30%) > fissures (11.74%). The average contribution rates of potential sediment sources in the basin to the suspended sediment at the outlet of the underground river were ranked as carbonate rocks (36.14%) > fissures (24.02%) > channels (22.93%) > clastic rocks (16.91%). There were totally different characteristics of soil erosion in the karst areas from those non-karst ones. In addition to the surface soil erosion and channel erosion, there was also soil leakage. The goodness of fit was greater than 80% for the surface and underground suspended sediment sources in the small watershed of Qingmuguan karst trough valley. Consequently, the credible model and composite fingerprint can be expected to effectively identify the main sources of the suspended sediment on the surface of the karst trough valley and the outlet of underground rivers. The finding can also provide a strong reference for the comprehensive management of small watersheds in the karst troughs.