| 06 Mar 2026
Impacts of Cascading Check Dams on Sediment Yield in the Middle Yellow River Basin: Insights from 50 Years of Grid-cell-level Simulation
Abstract. Check dams, globally built for controlling soil erosion, form complex cascading systems that pose significant challenges for assessing spatiotemporal dynamics of sediment yield (SY) at large basin scale. This study proposed an integrative framework combining dynamic sediment trapping efficiency of cascading check dams with the Revised Universal Soil Loss Equation (RUSLE), index of connectivity (IC), and sediment delivery ratio (SDR). This model was applied to evaluate grid-cell-based distribution of SY and sediment trapped by check dams during 1970–2020 in the middle Yellow River Basin (with over 47000 check dams). The Nash-Sutcliffe efficiency of proposed model increased to 0.71 compared to model ignoring sediment trapping of check dams (0.59). Check dams reduced the multi-year average SY by 50.01 % in dam-controlled areas. Totally 3.84 × 109 t of sediment was trapped over the 50 years, constituting 41.49 % of designed storage capacity. The sediment reduction contribution by check dams (SRCdam) exhibited considerable spatial heterogeneity, ranging from 73.9 % to 0.9 % among sub-basins, and the proportion of accumulated sediment to storage capacity of check dams (SARdam) varied from 78.1 % to 1.1 %. The SRCdam increased linearly with check dam density and the share of area they controlled, whereas SARdam increased logarithmically with SY from upstream of the check dams (P < 0.001). A trade-off between SRCdam and SARdam in some sub-basins indicates that the number of check dams in these basins is insufficient or overmuch. This study provides a practical and data-efficient method for assessing sediment trapping and reduction by cascading check dam systems in large basins, offering valuable insights for improving soil and water conservation strategies in erosion-prone regions.
General comments:
The manuscript “Impacts of Cascading Check Dams on Sediment Yield in the Middle Yellow River Basin: Insights from 50 Years of Grid-cell-level Simulation” presents an interesting attempt to integrate the RUSLE-IC-SDR framework with a dynamic sediment trapping efficiency (TE) module to evaluate the spatiotemporal impacts of cascading check dam networks on sediment yield. The topic is highly relevant to current challenges in large-scale soil erosion modeling and sustainable watershed management, particularly in the Middle Yellow River Basin. The authors’ effort to quantify the sediment reduction contributions and storage capacity dynamics of over 47,000 check dams using a novel topologic routing algorithm provides practical insights for optimizing soil and water conservation strategies and regional spatial planning. However, the manuscript in its current form suffers from several methodological uncertainties—particularly regarding spatial data resampling, parameter sensitivity, and the clarity of certain visual representations—that must be addressed.
Additional line comments:
Lines 24-27: The abstract mentions the results of the correlation analysis (such as the logarithmic growth relationship and P < 0.001), but does not mention specific R2 or correlation coefficients. I recommend briefly adding key statistics in the abstract to enhance the quantitative persuasiveness of the findings.
Lines 60-65: The introduction summarizes the shortcomings of existing research very well. It is recommended to further emphasize the specific advantages of the topological sorting algorithm proposed in this study in terms of “computational efficiency” and “avoiding double-counting” compared to traditional hydrological response unit or highly parameterized models.
Lines 126-128: The authors spliced and downscaled two NDVI datasets with significantly different spatial resolutions (8 km for AVHRR and 250 m for MODIS). This is an important source of uncertainty in the model. Please briefly add the validation accuracy of the downscaling technique (such as RMSE) to prove the reliability of the C factor calculation between 1970 and 2000.
Lines 165-171: Spatial data was resampled to a 100 m grid. Since topographic factors (LS factor) are highly sensitive to DEM resolution, resampling from 30 m to 100 m may flatten the slope, thereby underestimating erosion. Please explain the trade-off for choosing the 100 m resolution (e.g., computational efficiency) and discuss its potential impact on the final erosion estimation in the discussion section.
Lines 192-196: For the fine-grained sediment in the Loess region, the authors used an empirical coefficient of D = 0.046. Although literature is cited, it is recommended to discuss whether this parameter is time-variant under extreme rainstorm events on the Loess Plateau, and whether using a fixed parameter might overestimate or underestimate the trapping rate in specific years.
Lines 358-362: The discussion mentions the contribution of the “Grain for Green” program to vegetation coverage and erosion reduction. However, in the methodology section (Lines 163-164), the P factor (soil conservation practice factor) is assigned empirical values based on land use types. Please clarify whether the P factor in the 5 time slices from 1970 to 2020 can fully capture the dynamic changes brought by the Grain for Green program.
Table 1: The table lists the spatial resolutions of the original data (e.g., 250 m for SoilGrids, 30 m for DEM). I suggest adding a column in the table or clearly stating in the table footnote the “uniform resolution used for model calculation (100 m)” to improve the transparency of the methodology.
Figure 1: The numeric labels (1-17) for each sub-basin do not have enough contrast against some background colors (e.g., Sub-basins 14 and 15). I recommend bolding the numbers or adding a white halo effect to improve readability.
Figure 2: Typo correction: A close inspection reveals a spelling error in the red text description of “Step 2”. “RUSLR-IC-SDR” should be corrected to “RUSLE-IC-SDR”. Additionally, please briefly label the data transfer format next to the flow arrows in the figure to allow readers to intuitively understand the computational logic of the grid processing.
Figure 5: There are several black data points that significantly deviate from the fitted line in the scatter plot. Although the figure caption states that these points were excluded from the validation statistics, please add a sentence or two in the main text (Results section) explaining why these points were excluded (e.g., due to extreme flood events, missing data records, or anomalies caused by artificial water and sediment regulation).
Figure 11: The concept of the x-axis “Relative benefit for SRCdam, Relative benefit for SARdam” is not fully explained in the figure caption. Please provide a brief quantitative definition or a reading guide in the caption (e.g., explaining the physical meaning of a bar leaning to the left with a red outline) to help readers quickly grasp the core conclusion without having to refer strictly to the supplementary material (Text S2).