Zhang et al. (2025) Projections of actual and potential evapotranspiration from downscaled high-resolution CMIP6 climate simulations in Australia

Identification

• Journal: Hydrology and earth system sciences
• Year: 2025
• Date: 2025-12-01
• Authors: Hong Zhang, Sarah Chapman, Ralph Trancoso, Rohan Eccles, Jozef Syktus, Nathan Toombs
• DOI: 10.5194/hess-29-6863-2025

Research Groups

• Climate Projections and Services, Queensland Treasury, Queensland Government, Brisbane, Australia
• School of the Environment, The University of Queensland, Brisbane, Australia

Short Summary

This study evaluates actual and potential evapotranspiration (AET and PET) projections for Australia using dynamically downscaled high-resolution CMIP6 climate simulations, finding that these models provide reasonably accurate estimations and project scenario-dependent changes with significant implications for water security and agriculture.

Objective

• To evaluate the performance of AET estimates from observational-based products and the QldFCP-2 dataset for the historical period.
• To bias correct the QldFCP-2 AET dataset based on the best performing AET observational product and assess the effects of bias correction on mean climate and projected changes.
• To assess the climate change impacts on AET and PET in Australia across eight Natural Resource Management (NRM) regions.
• To investigate the regional drivers for changes in AET and PET.

Study Configuration

• Spatial Scale: Continental Australia, with analysis focused on eight Natural Resource Management (NRM) regions, at a 10 km spatial resolution. Validation performed at 26 OzFlux tower sites.
• Temporal Scale: Daily time step. Historical period (1960–2100, with specific evaluation for 1981–2010, and bias correction training for 1981–2020). Future projections for 2015–2100, with end-of-century changes assessed for 2080–2099 relative to 1995–2014.

Methodology and Data

• Models used:
  • Morton's Complementary Relationship Areal Evapotranspiration (CRAE) Model for daily AET and PET calculations.
  • Conformal Cubic Atmospheric Model (CCAM) for dynamic downscaling of CMIP6 Global Climate Models (GCMs).
  • Quantile Delta Mapping (QDM) for bias correction of derived AET.
  • Random Forest algorithm for identifying major drivers of AET and PET changes.
• Data sources:
  • Dynamically downscaled CMIP6 climate simulations (QldFCP-2 dataset), comprising an ensemble of 15 CMIP6 models (60 simulations across historical and three Shared Socioeconomic Pathway scenarios: SSP126, SSP245, SSP370).
  • Measured AET from 26 OzFlux eddy covariance flux tower sites in Australia for validation.
  • Seven gridded observation-based AET products for comparison: CMRSET AET, Australian Water Outlook (AWO), SILO, Copernicus AET, Global Land Evaporation Amsterdam Model (GLEAM), Derived Optimal Linear Combination AET (DOLCE), and ERA5-Land reanalysis.
  • Input variables for AET/PET calculations from QldFCP-2: daily maximum and minimum temperature, rainfall, solar radiation, vapour pressure, mean sea level pressure, leaf area index, and wind speed.

Main Results

• The QldFCP-2 ensemble mean AET estimates showed a mean error of 17% for the historical period (1981–2010) when compared against OzFlux data, which was comparable to or better than most observation-based products (ranging from 15.7% for CMRSET to 44% for SILO).
• Bias correction of QldFCP-2 AET using the AWO dataset as a reference did not consistently improve agreement with OzFlux data, increasing the overall mean percentage error from 17% to 22.3%.
• End-of-century (2080–2099 relative to 1995–2014) annual average AET projections for Australia indicate a decrease of -4.5% for SSP126 and -3.5% for SSP245, while an increase of 1.8% is projected for SSP370.
• In contrast, annual average PET is projected to increase across all scenarios: 5.0% for SSP126, 8.4% for SSP245, and 11.5% for SSP370.
• Regionally, AET changes are scenario and season dependent, with increases projected in northern and eastern coastal areas (e.g., Wet Tropics, Monsoonal North) under higher emissions, and decreases in drier inland regions (e.g., Rangelands, Southern and South-Western Flatlands, Murray Basin) particularly in Austral winter (June–August).
• PET consistently shows widespread increases across all regions, scenarios, and seasons, with generally larger increases in Austral winter (June–August) than Austral summer (December–February).
• Random forest analysis identified precipitation and solar radiation as the primary controlling factors for AET changes, while solar radiation and maximum temperature were the most important drivers for PET changes across most regions.

Contributions

• This study provides the first comprehensive projections of actual and potential evapotranspiration for Australia using dynamically downscaled high-resolution CMIP6 climate simulations (QldFCP-2).
• It offers a rigorous evaluation of multiple gridded observation-based and model-derived AET products against a network of 26 OzFlux sites, highlighting the substantial uncertainties in existing AET estimates.
• The research delivers new insights into the future changes of AET and PET across Australia's diverse Natural Resource Management regions and identifies their key climatic drivers, which is crucial for regional adaptation planning.
• It demonstrates that bias correction of downscaled AET using certain gridded observational products may not necessarily improve accuracy when validated against high-quality point observations.
• The findings emphasize the importance of regional-scale projections for informing decision-making in water security, agriculture, and natural resource management, given the significant spatial variability across Australia.

Funding

• Queensland Government
• The University of Queensland

Citation

@article{Zhang2025Projections,
  author = {Zhang, Hong and Chapman, Sarah and Trancoso, Ralph and Eccles, Rohan and Syktus, Jozef and Toombs, Nathan},
  title = {Projections of actual and potential evapotranspiration from downscaled high-resolution CMIP6 climate simulations in Australia},
  journal = {Hydrology and earth system sciences},
  year = {2025},
  doi = {10.5194/hess-29-6863-2025},
  url = {https://doi.org/10.5194/hess-29-6863-2025}
}