Anatahan island is 9.5 km east–west by 3.5 km north–south and truncated by an elongate caldera 5 km east–west by 2.5 km north–south. A steep-walled pit crater ∼1 km across and ∼200 m deep occupies the eastern part of the caldera. The island is the summit region of a mostly submarine stratovolcano. The oldest subaerial rocks (stage 1) are exposed low on the outer flanks and in the caldera walls. These include thick (∼10 m) and thin (2–3 m) lava flows, well-indurated tuffs, and scoria units that make up the bulk of the island. Rock compositions range from basaltic andesite to dacite, and most are plagioclase-phyric. On the steep north and south flanks of the volcano, these rocks are cut by numerous east–west-oriented, few-hundred-m-long lineaments of undetermined origin. Indurated breccias unconformably overlie scarps cut into stage 1 units low on the south flank. Intermediate-age eruptive units (stage 2) include caldera-filling lava flows and pyroclastic deposits and, on the outer flanks, vents and valley-filling lava flows. The youngest pre-2003 volcanic unit on Anatahan (stage 3) is a hydromagmatic surge and fall deposit rich in accretionary lapilli. Prior to 2003, this unit was found over almost the entire island, and in many places original depositional surfaces and outcrops could be found in high-energy environments along the coast, indicating a young (but undetermined) age. During reconnaissance visits in 1990, 1992, 1994, and 2001, geothermal activity (fumaroles as well as pits with boiling, sediment-laden pools) was observed in the southern part of the pit crater.

In March and April 1990, increased local seismicity, a large regional earthquake, and reported increased fumarolic activity in the pit crater prompted evacuation of Anatahan village, at the west end of the island. Our first field investigation took place in late April 1990 to assess the level of volcanic unrest, conduct reconnaissance geological observations, collect rock and geothermal water samples, and set up a geophysical monitoring network. Results at this time were inconclusive with respect to determining whether the activity was anomalous. Water in some of the geothermal pits within the pit crater was boiling, and pH values as low as 0.7 were recorded in the field. An electronic distance measurement (EDM) network was installed, and over a ∼1-week period, up to 9 cm of extension occurred across some lines but not others. Seismicity was characterized by intermittent local earthquakes but no sustained swarms or tremor. A brief visit in June 1990 revealed that the shallow lake near the boiling pits was gone, but activity in the pits themselves was similar to that of April 1990. Only minor extension had occurred along a single EDM line since the previous visit, and no earthquakes >M2.5 occurred during the visit.

Subsequent 1- to 2-day visits occurred in October 1990, May 1992, May 1994, and June 2001. Activity within the geothermal pits was relatively constant during every visit, although during this 11-year period the level of the water in each pit decreased. In June 2001, a ∼50-m-wide region of mud pots and steaming ground in the central part of the geothermal area had developed. No geologic evidence, however, suggested that an eruption would occur < 2 years afterward. Most of the EDM lines showed slight extension between late 1990 and 1992, followed by very gradual contraction from 1992 to 2001. A more extensive seismic-monitoring system was installed on the Northern Mariana Islands during these visits, and it recorded a small seismic swarm at Anatahan from May to July 1993. The telemetry component of the seismic equipment broke prior to 2001 and had not been repaired by the time of the May 2003 eruption, so no precursory seismic data were recorded to indicate pre-eruption unrest.