Applicability And Uncertainty Analysis Of APEX For Simulating Soil Erosion

Soil erosion is a global ecological and environmental problem which threatens human survival and progress.Studies of soil erosion processes and preventive measures have become a major focus among researchers worldwide.The Huaihe River Basin,one of the seven major river basins in China,has a catchment area of 269,800 km².Soil erosion rate can be as high as 35 t·ha^-1·yr^-1 across nearly 22%of the basin area.Hydrologic/Water quality models are effective tools that can be used to study soil erosion processes quantitatively, design soil and water conservation practices,and optimize soil and water resource management.It is important to study soil erosion under different management practices using simulation models because long-term detailed rainfall,runoff,and erosion data are not available for Huaihe River Basin.This study was conducted to evaluate the applicability of the APEX model to Chinese watersheds.Three plots located in the Middle and Upland Huaihe River Basin were selected as study areas.The APEX model was subjected to sensitivity analysis,calibration and validation using daily runoff and sediment yield data collected from the three plots during 1982-1986.Scenario simulations were conducted for 25-year periods using the calibrated and validated model for evaluating soil and water conservation benefits.Moreover, uncertainty analysis was conducted using Monte Carlo method.The main conclusions are as follows:1) Sensitivity analysis（SA） was conducted for the surface runoff and soil erosion components using Sobol and Extended Fourier amplitude sensitivity test（Extended FAST）. A total of 13 parameters were selected for SA.The SA results of the Sobol method showed that moisture condition 2 curve number（CN₂）,curve number index coefficient（CNIC）, conservation practice factor（PEC）,and peak rate-rainfall energy adjustment factor（APM） were influential for sediment yield,and that CN₂ and CNIC were influential for surface runoff.The Extended FAST method had similar results with Sobol and was more efficient computationally.2) The Monte Carlo scheme was used for model calibration/optimization in this study. Results from using multi-objective functions were better than those from using single objective function.The criteria of R²≥0.5,EF≥0.4 and percentage error of mean under 25% were established to assess the model results.The single objective function method failed to meet all the evaluation criteria（PE,EF and R²） simultaneously in the most cases.Predicted average daily runoff and sediment yield were within 15%and 20%of the corresponding observed values for the 1982 calibration period.The EF values ranged from 0.48 to 0.89, and R² from 0.56 to 0.98.The predicted average daily runoff and sediment yield were within 20%and 25%of the corresponding observed values for the validation period of 1983～1986. The calibrated APEX model tracked the variability of daily runoff and sediment yield well for the validation period,with EF values ranging from 0.41 to 0.84 and R² from 0.55 to 0.85. The goodness-of-fit measures indicate that the APEX model explained variation in the observed data reasonably well.The results suggested that the APEX model is applicable at the site-scale.It can be a useful tool for simulating runoff and erosion under different management practices in the Middle and Upland Huaihe River Basin.It will be useful in simulating runoff and erosion for ungaged watersheds.3) The scenario analyses using calibrated and validated APEX model revealed that the woodland with horizontal-ditch reduced surface runoff by 37%and sediment yield by 89%. The mixed wood-grass with horizontal-terrace reduced surface runoff by 35%and sediment yield by 84%.The results indicated that the construction of horizontal-terrace and horizontal-ditch and reforestation or revegetation are very effective in controlling surface runoff and sediment yield.The results were irradiative for choosing adaptive land management practices in the Middle and Upland Huaihe River Basin to reduce soil and water loss.APEX is a useful tool for evaluating the benefits of long-term soil and water conservation under different scenarios.4) Uncertainties of predicted runoff and sediment yield associated with uncertainties in PEC,CN₂,CNIC and APM were analyzed using Monte Carlo and statistical methods.The reduction in confidence intervals of surface runoff and sediment yield resulting from uncertainties of PEC and CN₂ for EHC4 under scenario 1 and 2 were also analyzed.The 95% confidence interval bands of average annual runoff over the 25-year simulation period ranged from 64.5 to 69.5 mm·yr^-1 for EHC1,from 58.0 to 62.8 mm·yr^-1 for EHC2,and from 103.0 to 105.9 mm·yr^-1 for EHC4.The confidence interval bands of average annual sediment yield ranged from 3.7 to 4.0 t·ha^-1·yr^-1 for EHC1,from 2.0 to 2.2 t·ha^-1·yr^-1 for EHC2,and from 25.3 to 26.2 t·ha^-1·yr^-1 for EHC4.The 95%confidence interval for average annual runoff reduction was 31.6%～36.2%for scenario 1 and 34.3%～38.7%for scenario 2.The 95%confidence interval for average sediment yield reduction was 83.4%～84.4%for scenario 1 and 88.9%～89.6%for scenario 2.