Long-term sedimentary sequence research can reveal how human activities and climate interact to affect catchment vegetation, flooding, soil erosion, and sediment sources. In this study, a biomarker sediment fingerprinting technique based on n-alkanes was used to identify long timescale (decadal) sediment sources in a small agricultural catchment. However, the highly saline carbonate environment and bacterial and algal activities elevated the levels of even-chain n-alkanes in the sediments, leading to an obvious even-over-odd predominance of short and middle components (C15-C26). Therefore, by analyzing three odd, long-chain n-alkanes (C27, C29 and C31) in 27 source samples from cropland, gully, and steep slope areas and one sediment sequence (one cultivated horizon and 47 flood couplets), a composite fingerprinting method and genetic algorithm optimization were applied to find the optimal source contributions to sediments. The biomarker fingerprinting results demonstrated that the primary sediment source is gullies, followed by cropland and steep slope areas. The average median source contributions associated with 47 flood couples collected from sediment core samples ranged from 0±0.1% to 91.9±0.4% with an average of 45.0% for gullies, 0±0.4% to 95.6±1.6% with an average of 38.2% for cropland, and 0±2.1% to 60.7±0.4% with an average of 16.8% for steep slopes. However, because farmers were highly motivated to manage the cropland after the 1980s, over half the sediments were derived from cropland in the 1980s. Biomarkers have significant advantages in the identification of sediments derived from different landscape units (e.g., gully and steep slope areas), and n-alkanes have considerable potential in high-resolution research of environmental change based on soil erosion in the hilly Loess Plateau region.