Scaled to Size: Downscaling Climate Models in Hawaiʻi

 Jan 21, 2016    by Lauren R. Kaiser

From a scientific standpoint, Hawaiʻi is a unique location for climate science in the Pacific Island Region. Since climate change is already impacting island nations throughout the region, you could call them the ‘canaries in the coal mines’ that serve as a warning to other areas. To avoid becoming casualties of climate change, organizations such as the Pacific Islands Climate Science Center and the International Pacific Research Center sponsor research projects and conference meetings that encourage using modeling approaches to better understand the climate system and how it responds to human activity. Since smaller geographic features like the Hawaiian Islands and other Pacific nations are typically overlooked in global climate models due to the coarse grid resolution of these models, there is a need to create reliable regional models to better represent the impacts of climate change in the Pacific Region.

Hawaiʻi’s unique topography and sharp climate gradients create a challenge for climate modeling methods. With elevation changes from sea level to over 4,200 m and some of the most dramatic precipitation and temperature gradients found anywhere in the world, simplifying the local climate into a mathematical model becomes quite a formidable task. Given the small size of the Hawaiian Archipelago, which is less than 30,000 km2, it is necessary to proportionally downscale the Earth’s climate systems to a more appropriate spatial resolution of 3 km (coarse) or 1 km (fine) wide grid cells.

Downscaling the global climate model structure for the Hawaiʻi Regional Climate Model (HRCM). Domain 1 (D1), 2 (D2), and 3 (D3) have grid spacing of 15 km, 3 km, and 1 km, respectively. Photo: (Hamilton, 2014)

Two modeling approaches have attempted to accurately build a detailed regional climate model for the Hawaiian Islands. The first approach, a dynamical downscaling approach aims to use the physical aspects of the landscape, such as changes in elevation, to replicate local climate patterns. Another approach applies statistical downscaling methods that simulate the characteristic climate of Hawaiʻi. Both methods (see here for a good overview of dynamical and statistical downscaling methods) have successfully built models that represent the typical climate found in the islands at a relatively fine resolution and both models perform fairly well, but it couldn’t be that easy now could it…?

Temperature trends at both the global scale and in each of these regional models show similar trends: warming. No matter what scenario you run or which models you use, the general projections show rising temperatures in the future. Where these two regional models do not agree, however, is when it comes to projecting general changes in future precipitation trends. What we see from the statistical downscaling method is an overall drying trend and decrease in rainfall while the dynamic methods project that the wetter, windward sides of the islands will continue to get wetter due to enhanced rainfall. These differences are not merely a result of variability or uncertainty, but they also actually show two completely different projected futures for the Hawaiian Islands. While neither of these two methods are incorrect, nor show completely erroneous projections, their divergent results do cause concerns for end users and managers. Because Hawaiʻi largely depends on rainfall for ground water recharge, it is imperative to understand what impacts climate change will have on local precipitation.  Going forward, end users, researchers and managers do not know which projections to incorporate into other climate change research or how to sustainably plan and manage future water resources. To resolve this issue of high uncertainty, researchers are collaborating to combine the best of both methods to determine more realistic and reliable climate projections.


Lauren R. Kaiser is a Researcher for the USGS Pacific Island Ecosystems Research Center (PIERC) at the University of Hawai'i System.

NOTE: Comments will be visible to the public. Before commenting for the first time, please review the ECCF's Editorial Policy.


tklemm's picture

Hi Lauren, and thanks for this article! I'm always grateful when people write about downscaling. I've heard about the two methods you mention, and I wonder which one of the two is better in which circumstances. Does one produce more accurate results but the other one is cheaper and faster to run? If I had to choose between results from either for a region, which one should I pick?
Thanks a lot!

lkaiser's picture

Hi Toni - the reality is that either method of downscaling to such a fine scale of 3 km or 1 km takes considerable processing time. I honestly can't tell you which to pick to use for the region as neither approach is wrong per say. However, there is a current National Center for Atmospheric Research (NCAR) project going on now that aims to bring together the benefits of both of these methods to make one regional model for Hawai'i. The hope is that by taking the best parts of each of these approaches, we can produce a more accurate regional model to make future projections with for other research projects and management practices. Everyone here is excited to see the results and are hoping that they will start coming out soon!

Add comment

Log in to post comments

Rain clouds gather around mountain in the Ko‘olau Mountain Range on the windward side of O‘ahu. Photo courtesy of Pacific RISA.