为揭示鄂西南地区公路路基边坡的土壤侵蚀规律，更准确的进行公路路基边坡土壤侵蚀预测。选择湖北沪蓉西高速公路恩施至利川段为主研究区，并以湖北沪蓉西高速公路宜昌至长阳段、宁夏银（川）武（汉）高速公路固原立交段作为附属试验区，通过对天然降雨过程、土壤侵蚀过程、土壤含水量等项目的观测、数据分析和整理，并结合土壤渗透性、土壤有机质含量、土壤粒径分析等室内试验结果，采用通用土壤侵蚀方程（RUSLE）建立了路基边坡土壤侵蚀预报模型，并得出如下主要结论：（1）鄂西南路基边坡土壤侵蚀预测模型中降雨侵蚀力指标（R）的最佳计算组合为： （2）上式中E60可用次降雨量P代替，从而减少计算量，代替后计算公式为， （3）实测值表明，当地路基边坡三种典型土壤的土壤可蚀性因子（K）值分别为：K红砂壤=0.032，K黄壤=0.0205、K红壤=0.0318；（4）与实测值相比较，诺模方程计算值偏大。出现该问题的主要原因，可能是当地土壤“粉砂+极细砂”含量与美国土壤差别较大；（5）在计算当地路基边坡土壤可蚀性因子（K）值时，可先用诺模方程进行计算，再用公式 进行修正。但使用该方法时需慎重，在有条件的情况下最好用实测值；（6）坡长因子（L）可用公式 计算，但公路路基边坡中坡度一般超过RUSLE的适用范围，因此不能用原有方法对m值进行核定；（7）通过实测值计算，将m值核定为0.35。公路路基边坡坡长因子（L）可用 计算；（8）67%，100%这两个路基边坡常见的坡度超出了以往的研究范围，不能用已有坡度公式计算。实测结果表明，33%、67%和100%三个坡度的坡度因子（S）值分别是3.70、7.18和10.37；（9）作物覆盖与管理因子（C）与覆盖度（V）呈指数函数关系，建立C-V关系式为 。植被对路基边坡的水土流失防治效益十分明显，覆盖度达到30%时，土壤侵蚀比裸地减少一半；（10）为准确反映工程措施的水土保持作用，应增加一个比例系数（B），即A=RKLSCPB；（11）对多层拱形防护边坡，计算坡长因子（L）时，坡长值建议选择总坡长λ，不要分解计算；（12）六棱砖P值为0.297，拱形框架梁P值为0.367，而机械压实P值为0.903；（13）在对路基边坡进行土壤侵蚀预测时，所建模型精度在一定程度上取决于不同类型小区所占的比例，坡度较缓的边坡所占比例越大，模型精度就越高。但从总体上看，该模型精度可以满足一般水土保持工作的需要；（14）以次降雨为单位进行侵蚀预测时，模型精度仅为78.2%，在对整个观测期土壤侵蚀进行预测时，模型精度可达到95.4%。模型的预报精度与时间序列长短有关，时间序列越长，模型预报精度也越高；（15）在自然条件与鄂西南近似地区，可以应用本研究中各因子计算方法进行土壤侵蚀预测。在自然条件与鄂西南相差较大的地区，应通过实测资料，重点确定降雨侵蚀力指标和土壤可蚀性因子值；（16）模型各因子计算方法符合鄂西南路基边坡土壤侵蚀特性。

In order to make a study of rules of soil erosion for subgrade slope in southwest Hubei and have a precise prediction of soil erosion, this study takes the expressway from Enshi to Lichuan in southwest Hubei as main research region. The expressway from Yichang to Changyang in Hubei and the expressway from Tongxin to Guyuan in Ningxia were choused as subsidiary research region. The process of rainfall, soil erosion and soil moisture content was recorded and analyzed. Soil infatuation characteristics, soil organic matter content and soil particle size were determined. The main conclusions achieved are as follows:(1)The best form for calculation of rainfall erosive index(R) in southwest Hubei is: (2)E60 can be replaced by rainfall P，thereby reducing the amount of computation. The conversion formula is: (3)The values of soil erosibihty factor (K) of three typical soils on local subgrade slope are respectively 0.032, 0.0205 and 0.0318.(4)Compared with the measured value, calculated values by Normal Equation is too large. The main reasons could be that “silt sand and very fine Sand” of local soil content is different from that of the United States.(5)In the calculation of the local soil erosibihty factor（K） of subgrade slope, Normal Equation can be used firstly. Then the result should be modified by the equation . However, this method is not very precise. You’d better use measured value under permit condition.(6)Slope length factor（L） in RUSLE is calculated by the formula . But the grade of subgrade slope is out of extension of RUSLE. Thereby, original method can not be used to define the value of m.(7)Slope length factor（L） of subgrade can be calculated by formula .(8)The subgrade slopes with 67%and 100% have not studied in RUSLE. Slope factor(S) of them can't be calculated with existed formula. The measured data reveals that slope facto(S)r with three grade 33%、67%、 100% are 3.70、7.18 and 10.37 separately.(9) An exponentially function exists between crop cover factor (C) and coverage V; the functional relationship of C-V is . Vegetation has an obvious effect on controlling the soil erosion on subgrade slope. When the vegetation coverage reaches 30 percent, soil erosion reduces by half than the bare soil.(10)In order to express the effect of engineering measurement for soil and water conservation, a proportional coefficient (B) should be added. A=RKLSCPB.(11)When calculating Slope length factor (L) of multilayer Arch slope, the slope length should be chosen the total length λ.(12)The effect of preventing soil erosion of hexagonal brick is best. That of arched frame is middle. And that of compaction by machine is poor.(12)The P value of hexagonal brick is 0.297. The P value of arched frame is 0.367. The P value of compaction by machine is 0.903.(13)When predicting soil erosion for subgrade slope, the accuracy of model lies on proportion of districts with different types. The proportion of slow slope is greater, the higher model accuracy. But generally, the accuracy of the model can meet the needs of water and soil conservation.(14)When predict soil erosion with hypo-rainfall, the model accuracy is only 78.2%.The accuracy can reach 95.4% when predicting soil erosion during the entire period of observation. Model forecast accuracy is related to the length of test time series, the time series is longer, the higher the accuracy of forecasting models.(15)For the region that have similar natural conditions with southwest Hubei, methodology presented by this study can be applied to predict soil erosion. For the region that have different natural conditions with southwest Hubei, Rainfall erosive index(R) and soil erosibihty factor（K） should be modified by measured data.(16)The Method of calculating factors in model is consistent with characteristics of soil erosion for subgrade slope in southwest Hubei.