ABSTRACT
Whether channel erosion or topsoil erosion constitutes the dominant erosion process throughout in the hilly region of the Chinese Loess Plateau (CLP), which suffers perhaps the most severe soil erosion in the world, had been controversial for a long time. The present article attempts to use the mid-infrared (MIR) spectroscopy fingerprinting method to trace sediment sources within nine small catchments in the hilly region of the CLP. Two major categories of sediment sources are identified: channel sediment and topsoil. Sediments trapped by check dams are used as the final sediment transferred by soil erosion. Discriminant analysis shows that MIR spectroscopy can differentiate between the two kinds of source sediments very well. The contributions of channel sediment and topsoil to the total final sediment are quantified using partial least squares regression (PLSR) analyses of MIR spectra to compare the trapped sediment samples with experimental models. The results of the root mean square error of calibration, root mean square error of validation and coefficient of determination for 18 models all show that the MIR-PLSR models boast very high prediction abilities in the nine catchments. A comparison between the geochemical fingerprinting method and the MIR spectroscopy method in one catchment reveals that although the two methods agree well on the channel sediment contributions, the two methods produce a significant difference (R2â¯=â¯0.4). Overall, the MIR-PLSR results show that channel sediments contribute 19% to 66% of the total sediment with an average of 33⯱â¯16% in the nine small catchments. Our results indicate that although channel bank sediment is important, topsoil erosion is the predominant process in small dam-controlled catchments on the CLP. Furthermore, the MIR spectroscopy fingerprinting method can provide a useful, non-destructive, rapid and inexpensive tool for tracing sediment sources from different kinds of loess.
ABSTRACT
This paper aims to investigate the hydrological response of a large-scale (8973â¯km2) mountainous watershed to different rainstorm spatial patterns and reforestation. Based on 32â¯years of observations, measurements of 184 rainstorm events and 125 sediment-producing events with complete hydrographs were analyzed. The K-means clustering method was used to classify the spatial patterns of rainstorm events in accordance with their event-based spatial rainfall characteristics. The 184 rainstorm events were classified into four spatial patterns, among which the spatial features differ significantly: (I) Spatial Pattern I (SPI) includes rainstorms with a low amount of cumulative areal rainfall (27.4â¯mm), the highest spatial variability (0.986), and the highest frequency; (II) Spatial Pattern II (SPII) includes rainstorms of high spatial variability (0.759) and the largest amount of local maximum daily rainfall (106.8â¯mm); (III) Spatial Pattern III (SPIII) includes rainstorms with a medium amount of cumulative areal rainfall (58.7â¯mm) and low spatial variability (0.362); and (IV) Spatial Pattern IV (SPIV) includes rainstorms with the largest amount of cumulative areal rainfall (117.2â¯mm) and the lowest spatial variability (0.313). Vegetation cover in the upper Du watershed was significantly improved after the implementation of the Grain-for-Green project. The average area-specific sediment yields (SSY) for the four SPs were 15.4, 65.5, 55.8, and 286.2â¯tâ¯km-2 before reforestation and decreased to 6.0, 59.3, 43.7 and 89.9â¯tâ¯km-2, respectively, after reforestation. ANOVA (analysis of variance) indicated that reforestation resulted in a significant reduction in runoff coefficient under SPIII and SPIV and a significant reduction in SSY under SPI and SPIV. A hysteresis analysis suggested that the proportion of events with a clockwise loop increased from 64.9% before reforestation to 82.1% after reforestation and that complex loops became less common during 2000-2010 under SPIV, thereby implying a reduced sediment supply.
