My guess is that most of you have either seen graphics or articles showing rainfall outlooks for some specified period. I’ve posted these types of graphics in previous blog posts, but in case this is your first read on the Mauka Showers Blog, here are some examples. Many of these outlooks depict probabilities for above, near, or below normal precipitation. Sometimes these outlooks are site specific, like the long-lead outlooks for Hawaiʻi from the Climate Prediction Center (CPC). Others are maps covering a broad area, with color shades depicting probabilities.
Screen shot of the Hilo portion of CPC’s long-lead Hawaiian Islands outlook. Red box highlights the precipitation portion of the outlook for Hilo. Click on the link for the description of the forecast abbreviations.
Precipitation rate forecast for July-August-September 2026 from the June 2026 run of the North American Multi-Model Ensemble (NMME). Brown shades are areas where probabilities favor below normal precipitation rates. Green areas favor above normal precipitation rates.
Knowing whether or not your area will have above or below normal rainfall can be very useful, but what does it mean in terms of rainfall amounts? For example, in May, the National Weather Service (NWS) called for above normal rainfall in the Hawaiian Islands during the May through September dry season based on climate model data and the CPC forecast (below). How much rainfall is “above normal”?
Excerpt from the NWS Honolulu Forecast Office’s Media Advisory issued on May 28, 2026 covering the dry season outlook. The highlight was added to show the rainfall forecast portion of the text.
Climate forecasts often use terciles, which divides data into 3 equally likely bins, above normal, near normal, and below normal. If you know the rainfall distribution over time, you can calculate these terciles and determine what these bins are for your specific location. Unfortunately, when site specific forecasts are available for Hawaiʻi, they are usually limited to the four main airport data sites. One of the main reasons for this is because these sites have the most consistent long term data records in the state. However, because of wide spatial variations in climate conditions across the state, airport rainfall is often not representative of conditions where most people live, work, farm, and play. You would need to know the tercile cutoffs for your specific location to determine what “above normal” means at your hale.
While I can’t show you what the tercile cutoffs are for every location in the state, I can show you what these look like for a limited set of sites just to show how these values vary across the island chain. The maps below show four different locations in the four main counties in the state (not including Kalawao County). It includes the main airport climate site, plus three other locations representing both windward and leeward spots. The terciles plotted are for the July-August-September (JAS) period, or the remainder of the current dry season. I added a table below the maps to show the values in case the numbers on the maps are hard for you to see.
Each site has an upper tercile threshold, lower tercile threshold, and the station name at the bottom. Rainfall totals for JAS at or above the upper threshold is in above normal territory. Totals at or below the lower threshold are below normal, and totals between the two values are near normal. You can see when “above normal” rainfall is predicted, expected rainfall totals vary considerably across the state. Obviously, just four sites per county is still insufficient, but at least you get a better idea of what the qualitative forecast means for your area.
July-August-September rainfall tercile thresholds for selected sites on Kauaʻi. For each site, the top number is the above normal tercile threshold, followed by the below normal threshold and the station name. Totals between the upper and lower numbers are considered to be near normal.
Same as above, but for Oʻahu.
Same as above, but for Maui.
Same as above, but for the Big Island.
| Location | Lower Threshold (inches) | Upper Threshold (inches) |
| Kauaʻi | ||
| Hanalei | 15.63 | 20.50 |
| Līhuʻe Airport | 4.15 | 6.58 |
| Wailua Exp Stn | 14.68 | 17.36 |
| Waimea Heights | 1.26 | 2.06 |
| Oʻahu | ||
| ʻĀhuimanu Loop | 11.91 | 17.65 |
| Honolulu Airport | 0.83 | 1.73 |
| Mililani | 4.86 | 6.51 |
| Waiʻanae Kawiwi | 0.79 | 1.58 |
| Maui | ||
| Hāna Airport | 10.93 | 15.08 |
| Kahului Airport | 0.67 | 1.58 |
| Kula Branch Stn | 1.75 | 3.33 |
| Lahainaluna | 0.63 | 1.43 |
| Big Island | ||
| Hilo Airport | 19.95 | 30.15 |
| Kamuela 1 | 7.29 | 11.19 |
| Kapāpala Ranch | 6.38 | 12.67 |
| Kealakekua TF | 12.42 | 17.95 |
I should also mention some things about how the tercile thresholds were calculated. All of these terciles were based on monthly rainfall data available from NWS or Hawaiʻi Climate Data Portal (HCDP) sources. Outside of the four main airport sites, it is difficult to find locations with complete monthly data records. To fill these gaps, I used the month-year data from the HCDP. The terciles are for the data period from 1991 through 2025.
Another important thing to mention is that climate forecasts are based on very coarse models which cannot resolve the complex island terrain. As a result, the complex micro-climates are not represented so a forecasted tercile from a climate forecast may not necessarily apply to all locations in the state. Getting around this limitation requires techniques such as dynamical or statistical downscaling. While some groups have used these techniques in Hawaiʻi, these have been for limited experiments and projects. I’m not aware of any operational climate models applying downscaling techniques for Hawaiʻi. “Operational” here means supported and updated routinely, and accessible 24/7. The NMME outlook is an example of an operationally supported product.
Depending on the interest level, I may produce more tercile thresholds for other locations across the state and for more periods than just JAS. If so, I may set up a “resources” page where interested users can access the data. It does take time to do the data work-ups so it won’t happen overnight for sure.