This atlas provides some statistics on seasonal rainfall normals and anomalies, and the frequency and intensity of drought episodes, in many of the independent Pacific Island nations and adjacent northeastern Australia. The analyses are based on a recently compiled Pacific Basin rainfall dataset covering the 1955-96 period. Some of these statistics were developed originally as a response to a request by the United Nations Department of Humanitarian Affairs, under the topic "Disaster Mitigation: A Policy Guide for Disaster Mitigation in the Pacific Islands Region". The stations included in this study are listed in Table 1, and shown in Fig.1 . These include five northeastern Australian stations and four Hawaiian Island stations, the latter of whose prediction potential and climate responses to ENSO were addressed in Chu (1995) and Barnston and He (1996). Including the Hawaiian stations, there is a total of 66 stations. Fifteen of the 66 are U.S.-affiliated; 14 of those are north of 5.5N; i.e., all except Pago Pago, in American Samoa. Fifty-one of the 66 stations are non-U.S.-affiliated (all south of 5.5N, and all except for four Kiribati stations south of the equator). Because most of the stations are located on islands, the term "island stations", "island climates", or the like, will often be used in this atlas to describe the 66 locations collectively, despite some of them being on very large islands such as Papua New Guinea or Australia. Because some of the stations are very close to one another (such as in Hawaii and in parts of Fiji), certain stations are artificially separated from one another in the graphics shown here, in order that numerical labels at the station locations do not block one another. In Table 1 the displacements applied to these few stations are given.

The statistics presented here are expressed relative to the entire 1955-96 (42-year) study period rather than a 30-year period in conformity with World Meteorological Organization (WMO) protocol. This decision is based on the fact that the atlas is developed in the latter half of the 1990s, such that normals based on 1961-90 are already getting "old" and will soon be replaced by 1971-2000 normals. By using a longer and more up-to-date climatological base period, it is hoped that the results will remain representative for a longer period than if the WMO period had been used. Additionally, a longer period reflects a greater number of ENSO episodes, whose typical effects on precipitation are better described with the larger sample.

Briefly, the contents of the atlas are as follows (to be discussed in more detail later): First, basic climatological statistics are presented for each of the 12 overlapping 3-month "seasons" (Dec-Jan-Feb, Jan-Feb-Mar, ..., Nov-Dec-Jan). The median (50 percentile [denoted as "%ile"]) amount, the extreme (record) amounts, and a pair of intermediate locations in the distribution (the 25 and 75 %ile amounts) are indicated diagrammatically. Greater detail about the climatological rainfall distribution is then provided by tables that indicate rainfalls corresponding to each decile boundary (the 10 %ile, 20 %ile, ..., 90 %ile) as well as the highest and lowest amounts recorded over the 1955-96 period. These rainfalls are expressed in terms of the percentage of the median amount for the given station and season.

Next, running 3-month precipitation totals from all available years (1955-96) are examined against the background of their 1955-96 climatology. Plots of the raw rainfall amounts (mm) and their percentiles throughout the 1955-96 period are provided for each station. One type of analysis that these statistics make possible is a tally of the number of consecutive running 3-month periods (loosely referred to as "seasons") over which the total rainfall was equal to or less than a given median-and-below percentile. For example, it might be found that at a given station there were 9 cases of two consecutive seasons with rainfall at or below the 20 %ile, 4 cases of three consecutive seasons, etc. Such counts for several 50-and-below percentile amounts are presented to characterize the nature of drought for each station.

The influence of ENSO episodes on Pacific basin rainfall is addressed using correlation analysis as well as using composite analysis in which the years are stratified in terms of their ENSO condition. The ENSO condition is defined here on the basis of sea-surface temperature and sea level pressure data in the tropical Pacific basin. Based on a number of other studies (Ropelewski and Halpert 1987, 1996; Lander 1994; He and Barnston 1996; Barnston and He 1996), the ENSO state is expected to shift the rainfall distribution noticeably, if not strongly, at many of the island locations. Examination on the individual station level provides greater detail than that found in the above-mentioned studies.

Finally, maps showing the spatial distribution of rainfall percentile for four regular (quarterly) seasons are presented for each year from 1955 to 1996. These provide some idea of the spatial coherence of rainfall anomalies using this particular set of Pacific basin stations. They also roughly illustrate the influence of the ENSO state on an individual case basis.

The atlas contents mentioned briefly in the four preceding paragraphs are discussed in more detail in sections 3 through 7 below.

Tables of the raw rainfall totals for each station and season are available on the Internet, at location http://www.cpc.ncep.noaa.gov/pacdir These data are provided for users who may wish to update our analyses after appending more recent data or to conduct their own rainfall analyses.

A future goal of this work is to identify at what point a drought condition is implied at a given locality. In view of the varying agricultural and hydrological needs among the islands, conditions defining a drought may be expected to differ from island to island. Drought-producing conditions may also change over time at a given island, based on changes in demand and the caliber of the water processing equipment and facilities (water distribution capacity, for example). The histories of the years that were considered to represent mild or serious droughts on the islands need to be examined in conjunction with the rainfall statistics to calibrate the latter for the individual economies and the possibly temporally changing levels of water demand and water processing capability. If useful, the conclusions of such a study may be published in a forthcoming journal note. This atlas presents the rainfall statistics that would be used as a basis for the estimation of drought thresholds, and the general social/economic consequences of drought, throughout the Pacific Basin.