Future projections of temperature change scenarios Downscaling of Temperature Changes for the Hawaiian Islands using CMIP5 Model Scenarios (delta-t)

The goal of this downscaling project was to develop an ensemble of future temperature change maps for the main Hawaiian Islands. At the time of this study, we did not have a sufficient number of station-based temperature data to achieve a good representation of the spatial structure of observed temperature variability and trends.  Thus it was difficult apply traditional statistical downscaling methods for the purpose of filling in detailed spatial information for future temperature change scenarios. Instead of using the observation-based statistical downscaling this method makes use of a regional climate model simulation as a source of the regional temperature information. In this study, the regional temperature pattern was obtained from simulations conducted with the Hawai‘i Regional Climate Model (HRCM). The model is a nested version of the advanced Weather Research and Forecasting (WRF) model with an inner domain resolution of 3 km (Elison Timm, 2017; Zhang et al., 2012, 2016a, 2016b). From the model-simulated changes a statistical relationship was obtained that quantifies how the projected temperature change varies with surface elevation height. This relationship was then used to scale the regional temperature change pattern over the islands proportional to the representative temperature change simulated by the individual CMIP5 models. The final temperature anomaly maps were obtained using a high-resolution digital elevation map: the elevation-dependent warming was mapped at a resolution of approximately 100m x 100m (longitude-latitude coordinate system).

Four different scenarios are available: two Representative Concentration Pathways (RCPs) scenarios RCP4.5 and RCP8.5, and for each the average statistics for the years 2040-2069 and 2070-2099.

The data products contain the ensemble mean of the 30-year average temperature change and measures of uncertainty due to parameter uncertainty in the statistically derived elevation-temperature relationship. The uncertainty from the ensemble member variability is provided for each of the four scenarios, too. Furthermore, the combined error is used to estimate the lower and upper confidence ranges (the mean +/- 2* standard error of the mean).

Note that this uncertainty does not include uncertainty in spatial pattern variations that are unaccounted for (‘unexplained variance’) in the elevation-temperature relationship.

Key to the application of this type of first-order downscaling process was that the elevation-dependent warming amplification factor is independent of the climate change scenarios (regarding scenario, timing, and climate sensitivity of the CMIP5 model members). Further details can be found in Elison Timm (2017)