Abstract: The Tianchi volcano in Changbai Mountains is a giant active volcano with potential risk of eruption, so more detailed researches of deep magma chamber as well as its closely related three-dimensional electric structure model are being concerned by many scholars. Based on adaptive regularized three-dimensional quasi-Newton inversion program developed by the author, this paper carries out three-dimensional magnetotelluric inversion including topography using the existing data observed decades ago. Our three-dimensional magnetotelluric adaptive regularization quasi-Newton inversion method improves the conventional quasi-Newton method by approximating Hessian matrix of the data misfit using LBFGS formula instead of total Hessian matrix which is approximated in conventional quasi-Newton inversion. This not only guarantees the precise regularization-term Hessian matrix, but also allows the regularization parameter to be adaptively updated on every inversion iteration without destroying the descent of total objective function. Thus, the regularization parameter's adaptive updating strategy on every iteration can be established based on L2-norm ratio of the data misfit function and regularization function, and our AR-QN inversion algorithm was implemented. The model synthetic data inversion results indicate that AR-QN inversion has very stable iteration flow, and is weakly dependent on initial model selection, which tends to be more suitable for the three-dimensional inversion task when MT survey sites are sparsely distributed on the surface.
Besides the impedances data, the tipper data which is more sensitive to horizontal conductive discontinuity was fitted in this three-dimensional inversion as well. Through comparison of qualitative analysis to measured data and anomalous-body-erasing forward modeling test, our new three-dimensional electric structure was proved to be a reliable model under the constraints of current magnetotelluric data. The three-dimensional electric structure is more accurate and rational compared with the electric structure obtained by previous researches based on two-dimensional inversion, and clearly characterizes the distribution of crustal magma chamber and magma channels with high resolution, even inverted by the limited spare-sites-distributed magnetotelluric dataset. Our result locates the crustal magma chamber on the northeast of Tianchi within the depths of 10~30km, and the general magma chamber along with magma channels system is represented as “V” shape. The magma channel of Tianchi volcano can be clearly shown on the EW profile 10km away from the north of Tianchi volcano. The magma chamber and channels system can be divided into three layers from shallow to deep: Magma channel with a depth of less than 5km and connecting to the Tianchi crater; alkaline fluid magma layer with a depth of 5~10km; and trachytic magma layer with a depth of 10~30km. As the coverage of the existing MT survey sites is still incomplete on the whole Tianchi volcano area, adding more MT survey sites in the northwest, southwest directions and especially in North Korea will help us get more reliable, higher-resolution three-dimensional electrical structure model for Changbaishan Tianchi volcano area.

Key words: Tianchi volcano in Changbai Mountains, electric structure, magnetotelluric, quasi-Newton inversion, adaptive regularization

摘要： 长白山天池火山是具有潜在喷发危险的巨型活火山, 对其深部低阻岩浆囊的精确探测和准确定位是人们关注的一个重要问题。 应用大地电磁方法研究深部电阻率结构是确定岩浆囊的最有效途径之一。 文中采用自主研发的三维大地电磁自适应正则化拟牛顿反演程序, 对长白山天池火山区大地电磁数据进行了带地形三维反演, 获取了较合理的深部三维电性结构模型。 反演除拟合大地电磁主阻抗外, 同时增加了对横向不均匀性更灵敏的倾子数据, 并通过定性分析对比、 模型验证的方法确认了三维反演的电性结构可靠性。 三维电性结构模型比以前的二维解释结果更加准确合理, 在有限稀疏的数据集下尽量精细地勾画了壳内低阻岩浆囊及通道的三维空间分布。 三维电性结构显示壳内岩浆囊位于天池东北部深10~30km处, 整体上岩浆囊和岩浆通道系统呈 “V”字形分布, 天池火山口的岩浆通道在天池以北10km的EW向剖面图上清晰可见。

关键词: 长白山天池火山, 电性结构, 大地电磁, 拟牛顿反演, 自适应正则化