Central Pacific Islands Downscaled Historical and Projected (1901-2100) Monthly Temperature and Precipitation Products

This set of files includes downscaled historical and projected monthly mean temperatures (in degrees Celsius, no unit conversion necessary) and monthly total precipitation (in millimeters, no unit conversion necessary) from Jan 1901 - Dec 2100 across the Pacific Islands region.

Included islands and associated spatial resolution of the downscaled data are as follows:

Commonwealth of the Northern Mariana Islands (90 Meters)
Guam (90 Meters)
Pohnpei, Micronesia (90 Meters)
Palau, Micronesia (90 Meters)
Kosrae, Micronesia (90 Meters)
Tutuila, American Samoa (90 Meters)
Manua, American Samoa (90 Meters)
Hawaii (440 meters)

Historical Models

CRU TS3.0 1901-2006
CRU TS3.1 1901-2009

Projections

2001-2100
Each set of files originates from one of five top ranked global circulation models from the CMIP3/AR4 models and SRES scenarios.

Projected Models

miub_echo_g - Meteorological Institute of the University of Bonn, ECHO-G Model
mri_cgcm2_3_2a - Meteorological Research Institute Coupled General Circulation Model version 2.3.2a
ipsl_cm4 - L’Institut Pierre-Simon Laplace Coupled Model version 4
mpi_echam5 - Max Planck Institute for Meteorology European Centre Hamburg Model 5
cccma_cgcm3_1 - Canadian Centre for Climate Modelling and Analysis (CCCma) Coupled General Circulation Model version 3.1 (CGCM3.1)

Note, cccma_cgcm3_1 did not have data for SRES A2 at the time of this analysis

The downscaling process utilizes PRISM climatological datasets from 1971-2000.

Methods

Data assemblage and standardization
SNAP obtained the required datasets to perform a model selection and subsequent downscaling procedure over the Pacific Islands region. This included temperature, precipitation, and sea level pressure across 1) 16 AR4 global climate model (GCM) datasets and 4 scenarios (20c3m, B1, A1B, A2), 2) the ERA40 reanalysis dataset covering 1958-2000, and 3) the 1971-2000 high resolution PRISM climatology datasets. All data was standardized into compatible formats, units, projections, and temporal and spatial resolutions for input into the model selection and downscaling procedures.

Model selection
Model selection methodologies closely followed Walsh et. al. 2008, but was customized for the Pacific Islands region by focusing on a bounding box encompassing all islands of interest. In general, the procedure compares monthly temperature, precipitation, and sea level pressure values at every 2.5 degree pixel location from 1958-2000 ERA40 outputs to the same variables output from 16 AR4 GCM twentieth century scenario runs (20c3m), calculates RMSE values and ranks them by a composite RMSE across all 3 variables. This resulted in 5 models being chosen as top candidates for downscaling.

miub_echo_g - Meteorological Institute of the University of Bonn, ECHO-G Model
mri_cgcm2_3_2a - Meteorological Research Institute Coupled General Circulation Model version 2.3.2a
ipsl_cm4 - L’Institut Pierre-Simon Laplace Coupled Model version 4
mpi_echam5 - Max Planck Institute for Meteorology European Centre Hamburg Model 5
cccma_cgcm3_1 - Canadian Centre for Climate Modelling and Analysis (CCCma) Coupled General Circulation Model version 3.1 (CGCM3.1)

Delta Downscaling
SNAP chose the island group of the Commonwealth of the Northern Mariana Islands as well as Pohnpei, Micronesia as case studies for downscaling. PRISM data was used as baseline climate, and because it was such high resolution data (440m for Hawaii, 90m for all other islands), and due to the fact that an individual island fell into a single GCM grid cell, the procedure was slightly modified to accommodate this large disparity and high resolution output. Outputs include temperature and precipitation from the top 5 models and three future climate scenarios (B1, A1B, A2) from 2001-2100. For more details about the delta downscaling method, please visit the methods section of snap.uaf.edu.