In 胆振地方 just before M6.7 in 2018/9/6, Distribution of Depth (brown line) was as below,
After M6.7, the new peaks has grown up around 30-40km depth mainly. Before M6.7, there were the even peaks of 30km/130km. By M6.7, the peak of deep area has been shifted to shallow area. M6.7's epicenter was in 40km. If the epicenter were shallower more, the curve might have become more simple like as trapezoid with downhill in 三陸沖. The even peaks of 30km/130km seems to be the trigger of avalanche from the deep area to shallow area.To make clear the relationship among the count changes per depth as time sequence, I tried making the graph of Counts of occurrence per depth for the data in 胆振地方 as below,
Note : There are not all data of each depth. Some depth data are omitted.
Intersections among the lines between of the different depths seem to be the changing points of the pattern of the count curves per depth. Counts of 22 times of 4 lines of 20km(orange)/30km(yellow green)/40km(green)/130km(gray) are crossed in narrow area around the time of M7.6. Towards Year 2000, there was the first jump of 20km curve, the 2nd jump of 10km, and 3rd jump of 1km. The jumps had moved to deeper area to shallower area. The close-sets between 2 lines seem to becoming the changing patterns as like circled points.
So, this graph may be useful for the prediction of the sudden increasing of counts (it may sometimes mean great earthquakes with many aftershocks). The long blank times until the next occurrence can be checked in the graph also.
In 三陸沖, the graph of counts of occurrence per depth is as below,
The graph is more simple because of less depths.Counts of 1km(red)/10km(yellow) have been overtaking and being overtaken. After jumping up, the line becomes flat, or tilted straight lines.The previous small jumping up were caused the next change in the other lines.