This study enlists rainfall observations from 1971-2017 for its analysis, which represents the start date of many stations in the PACRAIN database (www.pacrain.ou.edu). This allows for the maximum representation of rainfall pattern changes that have happened in this time, and also exceeds the minimum 30-year period recommended to account for climate variability cycles that could affect tropical Pacific Ocean rainfall. This study also uses strict data completeness protocols, such that only the gauges that collected daily rainfall observations on at least 99% of days from 1971-2017 are used. Furthermore, since this study is interested in rainfall patterns over the tropical Pacific Ocean itself, only rain gauges on low-elevation islands and atolls are enlisted for this study, by using a gauge elevation criterion of less than 200 meters. Another unique feature of this work is its combination of several variables that describe changes of rainfall that previous examples of research have used, but not in the same study. This includes standard indicators of changes of precipitation (in mm/yr) and frequency of precipitation (in days/yr) broken down by month and by year. However, these two indicators alone are not sufficient, since policy planners and governments are usually interested in extremities in precipitation, such as prolonged drought and rainfall during nth percentile rainfall events. As such, this study also considers changes in the intensity, frequency, and proportion of rainfall that occurs during extreme rainfall events (again separated by month and year), with extreme events defined as those that meet or exceed the 95th percentile at a given rain gauge. Additionally, trends in drought occurrence are analyzed by considering changes in monthly and annual maximum numbers of consecutive days with and without rainfall. Using a wide range of indicators allows for a holistic perspective on how rainfall patterns over the tropical Pacific Ocean have changed over the last few decades.
Annual results of this study are presented as spatial maps of the tropical Pacific Ocean, in which changes in rainfall pattern indicators are represented by triangles that are proportional to the magnitude of these changes. An example of these results is shown in Figure 1– trends in annual rainfall amount (in mm/yr) from 1978-1998. White triangles are indicative of trends that are statistically significant at the 90% confidence level, based on Spearman’s Rho regression analysis of a given rainfall indicator against the years 1978-1998 (the IPO’s positive phase) and 1999-2017 (the IPO’s negative phase) inclusive. These maps shall hopefully reveal how the trends in rainfall over this region have varied over the whole observational period and in accordance with changes in the IPO’s phase, which shall assist with attribution of the results of this study and understanding how the global warming hiatus has affected tropical Pacific Ocean rainfall patterns.
The results from these spatial maps have also been broken down by month (as mentioned above) and presented as scatterplots, in order to reveal a larger number of statistically significant changes in these variables that are perhaps negated by considering larger, i.e. annual, temporal scales. Further analysis of the statistically significant changes in the seven rainfall variables that are uncovered by this monthly breakdown reveals that all of the islands experience statistically significant changes in at least one variable and at least one month. An example of this is given in Figure 2 – the change in extreme rainfall proportion that occurred on Palau in July from 1971 to 2017. Based on the existence of statistically significant changes in the seven rainfall indicators used in this study, comparisons can be drawn with the results of previous studies that conducted related work. This shall be useful for validating this study’s results and potentially making attributions to climate change and relevant climate variability cycles.