Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Artificial Intelligence-Enhanced CMIP6 Climate Projections Across the Conterminous United States

Dataset ·
 [1];  [2];  [2]
  1. Oak Ridge National Laboratory; ORNL
  2. Oak Ridge National Laboratory

This dataset comprises high-resolution climate projections at 1/24 degree grid (~4km) over the conterminous United States (CONUS) based on ten Global Climate Models (GCMs) that are part of the Coupled Models Intercomparison Project phase 6 (CMIP6). The CMIP6 GCMs are downscaled using two artificial intelligence (AI) techniques, primarily based on the computer vision approach called super-resolution. We train two separate networks: super-resolution convolutional neural network (SRCNN) and super-resolution generative adversarial network (SRGAN). The networks are trained using Daymet observations, originally available at a 1 km resolution. For training purposes, the Daymet data is interpolated to 1/24 degree (~4km), 0.25 degree and 1 degree, which serve as high, intermediate and low-resolution inputs respectively. For each of the SRCNN and SRGAN network, we use a two-step resolution enhancement, the first step generates 4x refinement from 1 degree to 0.25 degree and the second step generates 6x refinement from 0.25 degree to 1/24 degree (~4km). We downscale daily scale precipitation, maximum temperature and minimum temperature for the six CMIP6 GCMs for 1980 to 2019 in the historical period and 2020 to 2059 in the near-term future under the shared socioeconomic pathway 585 and 245 (SSP585 and SSP245) emission scenarios. We also perform double bias-correction with Daymet observations using a quantile mapping approach, first for GCMs prior to making predictions at 1 degree grid and second after making final predictions at ~4km.

Research Organization:
Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States). Oak Ridge Leadership Computing Facility (OLCF); Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States)
Sponsoring Organization:
Laboratory Directed Research and Development (LDRD) Program, Oak Ridge National Laboratory; US DOE Office of Energy Efficiency and Renewable Energy (EERE), Renewable Power Office. Water Power Technologies Office
DOE Contract Number:
AC05-00OR22725
OSTI ID:
2530405
Country of Publication:
United States
Language:
English

References (1)

Daymet: Daily Surface Weather Data on a 1-km Grid for North America, Version 4 R1 collection January 2022

Similar Records

CMIP6-based Multi-model Hydroclimate Projection over the Conterminous US, Version 1.1
Dataset · Wed Feb 28 23:00:00 EST 2024 · OSTI ID:2311812

CMIP6-based Multi-model Hydroclimate Projection over the Conterminous US
Dataset · Thu Sep 01 00:00:00 EDT 2022 · OSTI ID:1887469

CMIP6-based Multi-model Streamflow Projections over the Conterminous US, Version 1.1
Dataset · Fri May 03 00:00:00 EDT 2024 · OSTI ID:2318650