A couple of months ago, while watching my weeds grow from all the rain, I did a post about the long term trends in wet season rainfall at selected sites across the state. I mentioned in the post that I thought the next step would be to do a post on dry season (May through September, a.k.a kau) trends. Well, that time is now. I looked at the same sites as I did for the wet season trends. Specifically, they are the four main airports, Līhuʻe, Honolulu, Kahului, and Hilo, and four other sites with long and generally reliable periods of record. These other sites, at Kōloa, Waiāhole, ʻUlupalakua Ranch, and Kapāpala Ranch, represent different climate regions from the airport sites. For reference, the eight sites are plotted on the map below.
Map of all the data sites used in this post. Locations are from the Hawaiʻi Climate Data Portal(HCDP).
Using the same methodology as the wet season trends plots, the graphs below show dry season rainfall from 1961 through 2025. Starting with the four main airports, the long term trend lines for Līhuʻe and Hilo show a slight drying trend over the 65-year period. However, neither of the plots pass the Mann-Kendall trend test (using the method at real-statistics.com) at the 0.05 significance level. Honolulu and Kahului are very flat, even though both have significant interannual variability. Looking at the running 10-year average trend lines, Hilo and Kahului show an interesting decadal signal. More on this later.
Plots of May through September (kau) rainfall totals for Līhuʻe Airport, Honolulu Airport, Kahului Airport, and Hilo Airport from 1961 through 2025. The red dashed line is a trendline based on the running 10-year average of the totals. The black dashed line is the linear trendline over the whole time series. The data are from the xmACIS2 database.
What about the four non-airport sites? These sites were chosen based on representation of different climate zones from the airport sites and their completeness of record. As with the wet season trends analysis, none of the four sites had perfect data records, so some gap filling was needed. As in the past, data from the HCDP month-year dataset were used to fill in the missing months.
In the graphs below, you can see that Waiāhole and Kapāpala Ranch show slightly increasing long term trends, Kōloa has a decreasing trend, and ʻUlupalakua Ranch is essentially flat. Only Kōloa’s decreasing trend passes the Mann-Kendall trend test at the 0.05 significance level. The four sites also show decadal trends, though Waiāhole’s trend is muted compared to the other three sites.
Same as the above graphs, but for Kōloa, Waiāhole, ʻUlupalakua Ranch, and Kapāpala Ranch. The map above shows the locations of these sites.
The long term trends generally agree with the findings in the Frazier and Giambelluca (2017) paper. The notable exceptions are at ʻUlupalakua Ranch and Kapāpala Ranch. Frazier and Giambelluca’s analysis shows statistically significant long term downward trends at both sites, whereas my data set does not. The differences could be due to a different definition of dry season (May through September in mine vs. May through October in theirs) and differences in period of record.
As mentioned above, the decadal cycle seen on several of the 10-year running mean trend lines is also interesting. The presence of decadal cycles in Hawaiʻi rainfall is well known (e.g. Chu and Chen 2005, and Frazier and Giambelluca 2017). However, the reason for the dry season cycle does not appear to be well known, or at least I haven’t seen or heard of a conclusive reason for the cycle.
Another interesting aspect of the decadal cycle is that it looks similar to the decadal cycle in annual tropical cyclone (TC) totals for the Central North Pacific (CNP). I’ve shown the following graph in other posts, but this time I added the 10-year running mean trend line. The timing of the maxima and minima in annual TC totals line up well with the maxima and minima in the 10-year trend line for rainfall. Note that the CNP TC numbers are for the entire basin (west of 140W longitude to the International Dateline, and north of the Equator). Therefore, peaks in the TC totals don’t necessarily mean more impacts, directly or even remnants, to the main Hawaiian Islands. There can be active TC years but most of the systems pass well south or north of the island chain without directly influencing rainfall. It could be that the broader scale cycle making TC activity favorable or unfavorable also helps regulate rainfall in Hawaiʻi. I don’t know if it’s related to the Pacific Decadal Oscillation and I’m honestly not familiar enough with all the research on the topic. It could be that some researcher has already published some papers on it, but I haven’t seen it. Or maybe I just forgot about the papers. After all, sometimes I forget why I went down the hall! Anyway, it would be really interesting to find out why these cycles line up.
Graph of the annual number of tropical cyclones in the CNP from 1970 through 2025. Red bars indicate El Niño during July-August-September (JAS), blue bars indicate La Niña, and yellow bars indicate ENSO neutral. Data provided by the Central Pacific Hurricane Center. Bar colors based on Relative ONI (RONI) data from CPC. Black dashed line is the running 10-year trend line.
I think that about does it for seasonal rainfall trends for now, unless I find something else that looks interesting. And before I forget, have a happy and safe 4th of July!
References:
Chu, P.-S., and H. Chen, 2005: Interannual and interdecadal rainfall variations in the Hawaiian Islands. J. Climate, 18, 4796-4813.
Frazier, A.G., and T. Giambelluca, 2017: Spatial trend analysis of Hawaiian rainfall from 1920 to 2012. Int J Climatol, 37, 2522-2531.