Abstract Four mechanisms caused tephra fallout at Soufrière Hills Volcano, Montserrat, during the 1995–1999 period: explosive activity (mainly of Vulcanian type), dome collapses, ash-venting and phreatic explosions. The first two mechanisms contributed most of the tephra-fallout deposits (minimum total dense-rock equivalent volume of 23 × 10 6 m 3 ), which vary from massive to layered and represent the amalgamation of the deposits from a large numbers of events. The volume of co-pyroclastic-flow fallout tephra is in the range 4-16° of the associated pyroclastic flow deposits. Dome-collapse fallout tephra is characterized by ash particles generated by fragmentation in the pyroclastic flows and by elutriation of fines. Vulcanian fallout tephra is coarser grained, as it is formed by magma fragmentation in the conduit and by elutriation from the fountain-collapse flows and initial surges. Vulcanian fallout tephra is typically polymodal, whereas dome-collapse fallout tephra is predominantly unimodal. Polymodality is attributed to: overlapping of fallout tephra of different types, premature fallout of fine particles, multiple tephra-fallout sources, and differences in density and grain-size distribution of different components. During both dome collapses and explosions, ash fell as aggregates of various sizes and types. Accretionary lapilli grain size is independent of their diameter and is characterized by multiple subpopulations with a main mode at 5ø. Satellite data indicate that very fine ash can stay in a volcanic cloud for several hours and show that exponential thinning rates observed in proximal areas cannot apply in distal areas.

Abstract The 26 December 1997 explosive activity of Soufrière Hills Volcano, Montserrat, provided an opportunity to study the evolution of a volcanic cloud by merging data from various satellites with wind-trajectory data. The activity involved a debris avalanche that descended SSW from the lava dome, to the coast, and a pyroclastic density current that traversed the coast and entered the sea. The slope failure and subsequent dome collapse occurred at c . 07:01 universal time (UT; 03:01 local time), lasted 15.2 minutes, and produced an upwardly convecting volcanic ash cloud that cloud temperatures suggest rose to c . 15 km. The volcanic ash cloud was unusual because the pyroclastic density current transported hot fine ash to the sea, where it rapidly transferred its heat to the sea water. The evaporation of large volumes of water produced a volcanogenic meteorological (VM) cloud that convected along with the volcanic ash cloud. The evolution of the volcanic and VM clouds was studied using an isentropic wind trajectory model and data from three satellite sensors: Geostationary Observational Environmental Satellite 8 (GOES 8), Advanced Very High Resolution Radiometer (AVHRR), and Total Ozone Mapping Spectrometer (TOMS). The high temporal resolution of the GOES 8 images filled many of the time gaps the other satellites left, and allowed quantitative retrievals to be performed using a two-band infrared retrieval method. The three-dimensional morphology of the volcanic cloud was reconstructed using GOES 8 data and by determining the heights of air parcels from wind-trajectory data. The volcanic cloud was estimated to contain up to 4.5×10 7 kg of silicate ash. Between c . 07:39 UT and 13:39 UT the ash signal of the volcanic cloud was masked by the VM cloud, which had a mass of up to 1.5×10 8 kg of ice. Ice forms when moist air is convected upwards to temperatures of less than −40°C and becomes saturated. Ice formation in volcanic clouds is especially likely when hot volcanic material is cooled by seawater rather than the atmosphere. The efficiency of evaporation of the seawater was calculated to be c . 5%, based on physical and GOES 8 data. TOMS data showed the SO 2 in the volcanic cloud rose higher than the ash in the volcanic cloud, as has occurred in several other eruptions. A comparison between GOES 8 and AVHRR data showed that AVHRR data retrieved higher fine-ash silicate masses and higher cloud areas than GOES 8 due to the finer spatial resolution of AVHRR images. The effect on retrieval data of the high water vapour content in the lower troposphere of the tropical atmosphere was quantified; the high humidity in the Montserrat region caused the characteristic ash signal to the infrared sensors to be depressed by up to 80%. This signal depression caused a corresponding underestimation of the mass and area of the volcanic cloud when the infrared brightness temperature difference retrieval technique was used.