A geometric technique is introduced to estimate the height of volcanic eruption columns using the generally discarded near-limb portion of geostationary imagery. Such oblique observations facilitate a height-by-angle estimation method by offering close-to-orthogonal side views of eruption columns protruding from the Earth ellipsoid. Coverage is restricted to daytime point estimates in the immediate vicinity of the vent, which nevertheless can provide complementary constraints on source conditions for the modeling of near-field plume evolution. The technique is best suited to strong eruption columns with minimal tilting in the radial direction. For weak eruptions with severely bent plumes or eruptions with expanded umbrella clouds the radial tilt/expansion has to be corrected for either visually or using ancillary wind profiles. Validation on a large set of mountain peaks indicates a typical height uncertainty of ±500 m for near-vertical eruption columns, which compares favorably with the accuracy of the common temperature method.

In a companion paper (Horváth et al., 2021), we introduced a new technique to estimate volcanic eruption column height from extremely oblique near-limb geostationary views. The current paper demonstrates and validates the technique in a number of recent eruptions, ranging from ones with weak columnar plumes to subplinian events with massive umbrella clouds and overshooting tops that penetrate the stratosphere. Due to its purely geometric nature, the new method is shown to be unaffected by the limitations of the traditional brightness temperature method, such as height underestimation in subpixel and semitransparent plumes, ambiguous solutions near the tropopause temperature inversion, or the lack of solutions in undercooled plumes. The side view height estimates were in good agreement with plume heights derived from ground-based video and satellite stereo observations, suggesting they can be a useful complement to established techniques.

We determine a 3-D P-wave velocity structure of the mantle down to 700 km depth under the Kamchatka peninsula using 678 P-wave arrival times collected from digital seismograms of 75 teleseismic events recorded by 15 portable seismic stations and 1 permanent station in Kamchatka. The subducting Pacific slab is imaged clearly that is visible in the upper mantle and extends below the 660-km discontinuity under southern Kamchatka, while it shortens toward the north and terminates near the Aleutian–Kamchatka junction. Low-velocity anomalies are visible beneath northern Kamchatka and under the junction, which are interpreted as asthenospheric flow. A gap model without remnant slab fragment is proposed to interpret the main feature of high-V anomalies. Combining our tomographic results with other geological and geophysical evidences, we consider that the slab loss may be induced by the friction with surrounding asthenosphere as the Pacific plate rotated clockwise at about 30 Ma ago, and then it was enlarged by the slab-edge pinch-off by the asthenospheric flow and the presence of Meiji seamounts. As a result, the slab loss and the subducted Meiji seamounts have jointly caused the Pacific plate to subduct under Kamchatka with a lower dip angle near the junction, which made the Sheveluch and Klyuchevskoy volcanoes shift westward.

A key problem of any video volcano surveillance network is an inconsistent quality and information value of the images obtained. To timely analyze the incoming data, they should be pre-filtered. Additionally, due to the continuous network operation and low shooting intervals, an operative visual analysis of the shots stream is quite difficult and requires the application of various computer algorithms. The article considers the parametric algorithms of image analysis developed by the authors for processing the shots of the volcanoes of Kamchatka. They allow automatically filtering the image flow generated by the surveillance network, highlighting those significant shots that will be further analyzed by volcanologists. A retrospective processing of the full image archive with the methods suggested helps to get a data set, labeled with different classes, for future neural network training.