Square Source Type Diagram
Abstract:Deformation in a small volume of earth interior is expressed by a symmetric moment tensor located on a point source. The tensor contains information of characteristic directions, source amplitude, and source types such as isotropic, double-couple, or compensated-linear-vector-dipole (CLVD). Although we often assume a double couple as the source type of an earthquake, significant non-double-couple component including isotropic component is often reported for induced earthquakes and volcanic earthquakes.
For discussions on source types including double-couple and non-double-couple components, it is helpful to display them using some visual diagrams. Since the information of source type has two degrees of freedom, it can be displayed onto a two-dimensional flat plane. Although the diagram developed by Hudson et al.  is popular, the trace corresponding to the mechanism combined by two mechanisms is not always a smooth line.
To overcome this problem, Chapman and Leaney  developed a new diagram. This diagram has an advantage that a straight line passing through the center corresponds to the mechanism obtained by a combination of an arbitrary mechanism and a double-couple [Tape and Tape, 2012], but this diagram has some difficulties in use. First, it is slightly difficult to produce the diagram because of its curved shape. Second, it is also difficult to read out the ratios among isotropic, double-couple, and CLVD components, which we want to obtain from the estimated moment tensors, because they do not appear directly on the horizontal or vertical axes.
In the present study, we developed another new square diagram that overcomes the difficulties of previous diagrams. This diagram is an orthogonal system of isotropic and deviatoric axes, so it is easy to get the ratios among isotropic, double-couple, and CLVD components. Our diagram has another advantage that the probability density is obtained simply from the area within the diagram if the probability density function of moment tensor’s eigenvalues P(λ1, λ2, λ3) depends only on the scalar moment [(λ12+λ22+λ32)/2]0.5. Even if this is not the real case, the easiness of calculating the areal density is useful when we compare the results of analyzing real data with that of analyzing background noise.