一种提高稀疏观测数据区反演精度的策略

A strategy to improve inversion precision in target area with sparsely distributed observation

在一个观测数据稀疏的地区,仅用观测数据来获得接近于真实的模型解是非常困难的,因此在反演时加入先验信息,考虑基于参考模型的反演方法是合理的。在一个观测数据有限,先验信息也很少的区域,获取合适的参考模型非常困难。在无法选取合适参考模型的情况下,提出了一种在参考模型远离真实模型的情况下,通过反演迭代过程中对参考模型的不断修改,使参考模型逐步逼近真实模型,从而提高了反演精度的策略。这里设计了均匀层状模型及含异常体的模型对算法进行了验证,证实了此方法对反演大型地质异常体具有较好的适用性。

Since the observed data of the target area are very sparse,it is difficult to obtain the model which is close to real model only using the observed data.Hence,it is reasonable to add some priori information into inversion and apply the inver-sion method based on reference model.With the lack of observed data but also and priori information,it is also very difficult to select an appropriate reference model.If an appropriate reference model can't be selected,how do we invert for a reliable solu-tion? This paper suggests a strategy to improve inversion precision,which encourages reference model near to real model by stepwise modifying reference model during inversion when the reference model is far away from real model.The authors de-signed a homogeneously stratified model and two models with the anomaly for testing algorithm.The inversion results illustrate that this strategy is appropriate for inversion of large size of anomaly.