中文摘要：

基于分形理论，选取黄土高原南部厢寺川林场地区20个典型样点，每个样点取3个剖面，共60个土样的实测土壤颗粒粒径分布数据，分别应用分形模型、对数正态模型、逻辑生长模型、WEIBULL模型预测土壤颗粒累积百分含量，提出一种预测土壤颗粒粒径分布的分形模型。结果表明，在0．002-0．1mm粒级范围内，分形模型对已知土壤资料的粒级个数和预测粒级的大小等因素并不敏感，具有较高的预测精度和稳定性：与对数正态模型、逻辑生长模型和WEIBULL模型相比，分形模型的总体预测误差最小且未出现大误差数据，可以有效对不同土粒分级标准间土壤质地资料进行转换。

英文摘要：

The soil texture is one of the most important indicators to reflect soil physical properties. It is the key input to many models, just as calculating the Erodibility K of the RUSLE and Pedo-Transfer Functions, which needs the soil texture of USA textural triangle. However, despite a number of recognized international standards, soil data are rarely compatible across national frontiers. Therefore, interpolation of the soil texture in different textural triangle is very necessary. Researches have shown that the soil has the fractal characteris- tic. In this study a fractal model is used for solving conversion of different soil texture triangle. For testing the stability and accuracy of the fractal model, 60 soil samples with different profiles and land-use were taken at Southern Loess Plateau. At first, 0.02mm particle data, 0.005mm particle data and omitting 0.02mm and 0.005mm particle data were omitted at the same time. Then the omitted particle data was predicted by the fractal model. The results indicate that the predicted particle-size data and the number of known particle-size data have little influence to the accuracy of the fractal model between the 0.002-0. 1mm soil fractions; it is demonstrate that the model is much better for predicting the particle-size data than Logistic growth model, WEIBULL model and Log-normal distribution model, the accuracy of the fractal model is satisfying and there are no significant errors about the predicted particle-size data. The fractal model can be used.for conversion of different soil texture triangle. More studies should be carried out.