ESPO-G6

Summary Description

ESPO-G6, the “Ensemble de Simulations Post-traitées d’Ouranos - modèles Globaux CMIP6” is an ensemble of bias-adjusted climate simulations created by Ouranos based on the global climate models (GCMs) from CMIP6. It downscales 3 variables for 26 global climate models over a historical period (1950–2014) and three future pathways (SSP2-4.5, SSP3-7.0, SSP5-8.5) using the Detrended Quantile Mapping (DQM) bias-adjustment method. A detailed description of the method is available in Lavoie et al. (2024). There are two versions of the of ESPO based on two reference datasets: ESPO-G6-R2 (RDRS/CaSR v2.1) and ESPO-G6-E5L (ERA5-Land). ESPO-G6-R2 is recommended by Ouranos in most cases and serves as the database for Ouranos Climate Portraits.

ESPO-G6 dataset animation

An animated illustration of 30 year temperature anomalies calulated from ESPO-G6 for Quebec and Montreal compared to the 1991-2020 reference period (source: Ouranos)

Dataset Characteristics

Strengths and Limitations

Key Strengths of ESPO-G6

Strength Description
Accessibility Climate projections from this dataset can easily be explored through the online platform Portraits Climatiques, and many relevant indicators are already computed.
Multi-Model Ensemble One member of each available GCM was adjusted to give all models equal representation.
Multiple emission scenarios. The ensemble includes simulations driven by three different emission scenarios, which allows it to span a wide range of possible futures.

Key Limitations of ESPO-G6

Limitation Description
Biases remain The DQM method does not specifically adjust for sequences, extremes and spatial patterns.
Limited variables Focuses primarily on temperature and precipitation; other climate variables are not included.

Expert Guidance

CMIP6 raw simulations have been bias-adjusted (or statistically downscaled) to create ESPO-G6. Technically, this process should make the data more suited to be used in a decision-making context, as the resolution is increased and most biases have been removed. The goal is to use the climate of the reference with the climate change signal of the GCM. A good application for the dataset would be to calculate climatologies of climate indicators (ex. number of days above 30 degC, total accumulated precipitation), rather than values on individual days. If possible, it is ideal to look at the deltas of climatologies (difference between the present and future).

There are a few caveats that generally come with bias-adjustment. Those are listed on XX page [TODO]. Here are some considerations that are particular to the ESPO-G6 v1.1.0.0 method. The DQM method is used which means that the trend from the GCM is preserved and that the distribution from the reference is adopted (i.e. the mean, variance and quantile of the adjusted simulation will be the same as the one from the reference). However, the method is not designed to correct specifically sequences, extremes, multivariate relationships and spatial pattern. Further, the adjustment is applied on the daily temperature range, the daily maximum temperature and the precipitation. This ensures that there are no temperature inversion in data. However, it can mean that the variance of the daily minimum temperature (which is computed afterwards from the range and the maximum) can be underestimated (Lanzante et al. (2025)).

Choosing the models

Every CMIP6 GCM that was available on ESGF for the variable and scenarios of interest were bias-adjusted. However, in general, we do not recommend that all of them should be used if the timing of the warming is important for the application. Indeed, many CMIP6 models have a climate sensitivity that is much too high (Hausfather et al. (2022)). In order to avoid this “hot model problem”, only models with a Transient Climate Response (TCR) in the likely range (1.4–2.2 °C) should be used (Table 1).

Table 1. Members of ESPO-G6

Institution Model Member TCR (degC) In TCR likely range
CAS FGOALS-g3 r1i1p1f1 1.50 ✓
CMCC CMCC-ESM2 r1i1p1f1 1.92 ✓
CSIRO-ARCCSS ACCESS-CM2 r1i1p1f1 1.96 ✓
CSIRO ACCESS-ESM1-5 r1i1p1f1 1.97 ✓
DKRZ MPI-ESM1-2-HR r1i1p1f1 1,64 ✓
INM INM-CM5-0 r1i1p1f1 1.41 ✓
MIROC MIROC6 r1i1p1f1 1.55 ✓
MPI-M MPI-ESM1-2-LR r1i1p1f1 1.82 ✓
MRI MRI-ESM2-0 r1i1p1f1 1.67 ✓
NCC NorESM2-LM r1i1p1f1 1.49 ✓
CNRM-CERFACS CNRM-ESM2-1 r1i1p1f2 1.83 ✓
NIMS-KMA KACE-1-0-G r1i1p1f1 2.04 ✓
NOAA-GFDL GFDL-ESM4 r1i1p1f1 1.63 ✓
BCC BCC-CSM2-MR r1i1p1f1 1.55 ✓
MIROC MIROC-ES2L r1i1p1f2 1.49 ✓
CCCma CanESM5 r1i1p1f1 2.71
CNRM-CERFACS CNRM-CM6-1 r1i1p1f2 2.22
EC-Earth-Consortium EC-Earth3 r1i1p1f1 2.30
IPSL IPSL-CM6A-LR r1i1p1f1 2.35
MOHC UKESM1-0-LL r1i1p1f2 2.77
NCC NorESM2-MM r1i1p1f1 1.22
EC-Earth-Consortium EC-Earth3-CC r1i1p1f1 2.63
NUIST NESM3 r1i1p1f1 2.72
EC-Earth-Consortium EC-Earth3-Veg r1i1p1f1 2.66
INM INM-CM4-8 r1i1p1f1 1.30
AS-RCEC TaiESM1 r1i1p1f1 1.30

Choosing the scenario

“Ouranos suggests adopting a risk analysis perspective, and interpreting SSP2-4.5 as the median scenario and SSP3-7.0 as the upper scenario. According to Working Group III of the latest IPCC report, SSP5-8.5 (like RCP8.5) does not represent a typical business-as-usual scenario and is currently considered unlikely. It is useful only as a high-risk scenario, representative of the upper limit of “no climate policy” scenarios. SSP5-8.5 may be relevant in the context of unlikely hazards with catastrophic consequences, or as an analogue for a post-2100 climate, useful for land-use planning, for example.” (Source: Portrait Climatiques (2025) )

Choosing the reference

There are two sub-ensembles of ESPO-G6 based on two reference datasets: ESPO-G6-R2 (RDRS/CaSR v2.1) and ESPO-G6-E5L (ERA5-Land). Officially, Ouranos recommends ESPO-G6-R2. However, users with specific regions can consult the CaSR and ERA5-Land pages to see if a dataset is better suited to their region. If a user has the capacity to handle a large ensemble and wants to cover the full range of uncertainty, they could also use both sub-ensemble.

Problematic Areas

  • Users should be careful with precipitation data close to the south edge of the North American domain where there is less trust in the reference data.
  • Some small regions in Alaska and Greenland showed very small tasmin and have been masked out by NaNs for 2 models (BCC-CSM2-MR and GFDL-ESM4 ). More details are available in section Health Checks of Lavoie et al. (2024).

Validation

A validation using the VALUE framework is presented in Lavoie et al. (2024).

Example Applications

links to Electricity Sector Activities - Electricity System Planning – Demand and generation forecasts under climate change
- Operations Planning – Assess risks to hydropower operations from variability and extremes
- Infrastructure Planning and Asset Management – Evaluate long-term exposure to extreme heat and drought for transmission lines and cooling water supply
- Assurance and Reporting – Provide climate risk evidence for ESG and regulatory disclosures

Available Variables in ESPO-G6

For details click on variable group to uncollapse

  • Daily maximum temperature [K]
  • Daily minimum temperature [K]
    (Calculated from adjusted tasmax and adjusted? daily temperature range)
  • Mean daily precipitation flux [kg m-2 s-1] (equal to [mm s-1])

Data Access

The ESPO-G6 dataset’s bias adjusted climate variables and all indicators calculated for Ouranos Climate Portraits are available on PAVICS for the North American continent. To avoid downloading very large datasets in their entirety PAVICS allows partial/regional extraction and provides tutorials to do so. With a free PAVICS user account, the Jupyter notebook with the Python code from the tutorials can be directly used on PAVICS. The THREDDS catalogue on PAVICS to access the ESPO-G6 dataset is available here. In all other dataset description we pointed to PAVICS, not to the THREDDS catalogue. Juliette proposed to do it this way. To be discussed. After discussion: Create a test case where we put both to see feedback. The Python code used in the generation of ESPO-G6 is available in this GitHub repository.

References

Hausfather, Z., Marvel, K., Schmidt, G.A., Nielsen-Gammon, J.W., Zelinka, M., 2022. Climate simulations: Recognize the ’hot model’ problem. Nature 605, 26–29. https://doi.org/10.1038/d41586-022-01192-2
Lanzante, J.R., Dixon, K.W., Adams-Smith, D., 2025. Variance distortion via indirect downscaling of daily minimum temperature from diurnal temperature range. Journal of Applied Meteorology and Climatology 64, 1093–1104. https://doi.org/10.1175/JAMC-D-25-0009.1
Lavoie, J., Bourgault, P., Logan, T., Caron, L.-P., Gammon, S., Smith, T.J., Biner, S., Braun, M., 2023. ESPO-G6-R2 : Ensemble de simulations post-traitées d’ouranos - modèles globaux CMIP6 - RDRS v2.1 / ouranos ensemble of bias-adjusted simulations - global models CMIP6 - RDRS v2.1. https://doi.org/10.5281/zenodo.7877330
Lavoie, J., Bourgault, P., Smith, T.J., Logan, T., Leduc, M., Caron, L.-P., Gammon, S., Braun, M., 2024. An ensemble of bias-adjusted CMIP6 climate simulations based on a high-resolution north american reanalysis. Scientific Data 11, 64. https://doi.org/10.1038/s41597-023-02855-z