中文摘要：

为了更好地进行土壤水分反演，发展了一种基于ALOS／PALSAR数据、利用自适应神经模糊推理系统（adap-tireneurofuzzyinferencesystem，ANFIS）反演土壤水分的方法。首先，根据研究区实际情况，利用AIEM和Oh模型模拟了试验区裸土区的后向散射特性，建立了后向散射系数与地表粗糙度之间的关系；然后，考虑到研究区地表粗糙度几乎没有变化这一情况，设定了地表粗糙度对后向散射系数的影响为常量；在此基础上，分别利用ANFIS，BP神经网络、多元线性回归和多元非线性回归方法构建了裸土区土壤水分的反演模型，并利用野外实测数据对模型进行了验证。研究结果表明，采用ANFIS方法构建的模型反演精度最高，其均方根误差为0．030，相对误差为14．5％。因此，可以利用ANFIS方法反演裸土区的土壤水分含量，其反演结果具有较高的精度。

英文摘要：

A new algorithm of soil moisture inversion was developed by using ANFIS based on ALOS/PALSAR data. Firstly, the surface backscattering characteristics in bare region were simulated and the relationship between the backscattering coefficient and the surface roughness was built in consideration of the actual situation. Secondly, the surface roughness in the study area didn＇ t change due to the fact that the effect of backscattering coefficient brought by surface roughness was constant. On such a basis, a model of soil moisture inversion was built by using such methods as ANFIS model, BP neutral network, multiple linear regressions and multiple nonlinear regressions, and then the simulating data was used to validate the accuracy of this model. The result shows that the estimated soil moisture error is 0. 030, and the relative error is 14.5%. Therefore, the ANFIS method can be used to conduct inversion of soil moisture in bare region with higher precision.