土壤传递函数模型的研究进展

摘要

研究评述了线性回归、非线性回归、神经网络等方法构建土壤传递函数模型的优缺点,介绍了传递函数模型及其应用不确定性分析中常用的Bootstrap analysis、改进的Monte Carlo、最小二乘法等方法,探讨了输入数据、模型结构和参数对传递函数及其应用不确定性的影响,并指出今后可借鉴分析法里的格林函数法和耦合去耦等方法开展土壤传递函数及其预测不确定性的研究,明确传递函数模型不确定性的主要来源,提高模型的应用效率.%This paper evaluated advantages and disadvantages of methods for PTFs development using linear regres-sion , nonlinear regression and artificial network; and then reviews the methods (I.e. Bootstrap analysis, modified Monte Carlo, least-squares optimization. ) to analysis soil pedotransfer function performance and its appliance uncertainty; finaly discussed the methods of Green' s function, coupled its decoupling may be introduced to further study on uncertainty of PTFs, and the main source of prediction uncertainty need be further affirmed in order to improve the model utilization ef-ficiency .