土层厚度是坡地水文过程的一个重要控制因素。使用瑞典MALA公司生产的ProEx系统探地雷达,分别采用100和500MHZ频率的天线对土层厚度进行探测,然后通过开挖探槽实测土层厚度,同时调查了影响探地雷达结果的因素——土石交界面处的基岩风化度,并建立了不同自变量下的线性回归模型和GEP模型,对探地雷达在喀斯特坡地土层厚度估测中的适用性进行了探讨。结果表明,喀斯特坡面土层浅薄且含有较多碎石,使用频率较高的天线可以提高探测精度;以100、500MHZ频率天线的探测结果和基岩风化度这3个自变量的组合作为输入变量,较仅使用一种频率天线的探测结果为输入变量的模拟精度高;相同输入变量条件下,受限于实际探测情况,基于GEP算法建立的土层厚度模型较经典统计的线性回归模型预测精度的提升空间有限,因此采用线性回归模型即可,以100、500MHZ频率天线的探测值和基岩风化度为自变量的线性模型的决定系数和均方根差分别能达No.660和15.0cm。
Thickness of the soil layer on a hillslope is an important factor governing hydrologic processes on the hillslope. A field survey was carried out using a ProEx ground penetrating radar ( GPR ) , product of MALA in Sweden, to detect thickness of the soil layer on a hillslope in a karst region. The detection used two antenna frequencies, 100 and 500 MHZ, separately, Then exploratory trenchess were excavated on the slope to determine thickness of the soil layer physically and at same time to explore for factors affecting accuracy of the radar detection, like weathering and the bedrock. On such as basis, a linear regr d es egree of the bedrock at the interface between the soil layer sion model and a GEP model was established with different independent variables to verify applicability of the ground penetrating radar to soil thickness detection on karst hillslopes. Results show that as in the karst region, the soil layers on hillslopes are generally thin and contain a lot of debris, the use of a higher frequency antenna may improve detection precision; the use of the combination of the results of the radar detections with two different frequencies and weathering degree of the bedrock as input variable is higher in simulation accuracy than the use of the results of the radar detection with a frequency as input variable. When input variables are the same, the GEP model is very limitedly higher than the linear regression model in prediction accuracy because of limitations in actual detection. So, the latter is recommanded. The linear regression model combining results of the detection using 100 and 500 MHZ antennas and weathering degree of the bedrock as independent variable may reach 0.660 and 15.0 cm in coefficient of determination and root mean square error, respectively.