CanDCS-M6

Summary Description

CanDCS-M6 is a Canada-wide, statistically downscaled CMIP6 dataset that provides daily precipitation, maximum temperature, and minimum temperature at ~10km resolution for the period 1950-2100. It downscales 26 global climate models over a historical period (1950–2014) and four future pathways (SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5) using the multivariate bias-correction method MBCn to preserve both marginal distributions and inter-variable dependence. The calibration target is a blended observational dataset referred to as PCIC-Blend, a variant of the NRCANmet dataset). Evaluation shows improved representation of multivariate and compound climate indices relative to previously used univariate method (CanDCS-U6), while univariate performance remains comparable. The dataset is used to project temperature- and precipitation-based indicators over Canada on the Canadian web portal ClimateData.ca. A detailed description of the dataset is available in Sobie et al. (2024).

Download interface for CanDCS-M6

Minimum temperature of Australian climate model ACCESS-CM2 in December 1999 - a view taken from the PCIC/UVic download interface for CanDCS-M6

Dataset Characteristics

When to use CanDCS-M6

  • When you need CMIP-based, bias-adjusted future climate projections for Canada
  • When you need downscaled climate data suitable for impact studies requiring higher spatial resolution than CMIP6 and that better represent local processes (e.g., topography, coastlines) than global models
  • When you need climate projections under four future emissions pathways (SSPs)
  • When you need scenario-based projections with improved agreement with the historical baseline of gridded observations dataset NRCANmet
  • When you need inputs for impact models that are sensitive to biases (e.g., hydrology, energy demand models)
  • When you want to assess threshold-based metrics (e.g., degree days, exceedances, return levels)
  • When you want to avoid performing bias correction and post-processinng workflows on CMIP6 data

Strengths and Limitations

Key Strengths of CanDCS-M6

Strength Description
Accessibility Climate projections from this dataset can easily be explored through the online platform ClimateData.ca, and many relevant indicators are already computed.
Regional model subsets A representative subset of models has been identified, both for Canada as a whole and five Canadian subregions.
Multiple emission scenarios The ensemble includes simulations driven by four different emission scenarios, which allows it to span a wide range of possible futures.
Large multi-model ensemble Makes use of a large number of climate models, which allows to sample the (structural) uncertainty associated with climate projections.
Effective for multivariate/compound indices The dataset reproduces multivariate/compound indices (e.g., hot-dry days, precipitation-as-snow) more effectively than datasets using an univariate bias adjustment, which could improve analyses where temperature–precipitation dependence matters.

Key Limitations of CanDCS-M6

Limitation Description
Limited variables Focuses primarily on temperature and precipitation; other climate variables are not included.
Includes “hot models” The ensemble includes a large number of models that have a climate sensitivity larger than the observed value, which leads them to warm up slightly faster (at the global level) than we would expect.

Expert Guidance

The authors of CanDCS-M6 recommend the use of their dataset for climate impact assessments, as well as hydrologic, agricultural, or snow modelling. They also note a few limitations. First, similarly to ESPO-G6, the trend of the GCMs is preserved in the bias-adjusted product. This means that “the hot model problem” (Hausfather et al. (2022)) will also be present for this dataset. However, the authors of the dataset provide no guidance on this issue. See the ‘Expert Guidance’ section of the CMIP6 dataset for more information.

Second, only one member is available for each GCM. While this can lead to an underestimation of the internal variability it is good practice in building multi model ensembles to give each model “one vote”. This produces a balanced ensemble which would not be the case if all member runs from models were included because not all models produce equal numbers of member simulations.

Third, the biases of the reference dataset will appear and are likely to create a smoothing, especially in region with less station data (e.g., in the north). In these areas, they recommend supplementing the analysis with other datasets, such as CORDEX RCMs or CanLead.

Note that the SSP5-8.5 scenario is available for this dataset. According to the recommendation of Working Group III, this scenario is currently considered unlikely.

The previous version of this dataset was called CanDCS-U6. It was using the BCCAQv2 bias-adjustment method with the NRCANmet reference dataset. A comparison of the two datasets by their authors shows an improvement on multivariate indices (due to the new method) and an improvement on precipitation on the west coast of Canada (due to the new reference dataset).

Variables available in CanDCS-M6

For details click on variable group to uncollapse

  • Daily maximum temperature (°C)
  • Daily minimum temperature (°C)
  • Daily total precipitation (mm)

Many climate indicators based on minimum temperature, maximum temperature and precipitation have been pre-calculated from the CanDCS-M6 dataset. See the download section for details.

Data Access

Climate information for Canada based on the CanDCS-M6 dataset can be visualized and browsed on ClimateData.ca. To access and use the dataset in other contexts it can be downloaded from PCIC/UVic’s download interface for CanDCS-M6. Indices derived from CanDCS-M6 for Canada can be downloaded from the respective portal provided by the government of Canada. Both, the downscaled climate variables and the indicators are also available on PAVICS. 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 in the tutorials can be directly used on PAVICS.

References

Hausfather, Z., Marvel, K., Schmidt, G. A., Nielsen-Gammon, J. W., & Zelinka, M. (2022). Climate simulations: Recognize the ’hot model’ problem. Nature, 605(7908), 26–29.
Sobie, S. R., Ouali, D., Curry, C. L., & Zwiers, F. W. (2024). Multivariate canadian downscaled climate scenarios for CMIP6 (CanDCS-M6). Geoscience Data Journal, 11(4), 806–824.
Werner, A. T., Schnorbus, M. A., Shrestha, R. R., … Anslow, F. (2019). A long-term, temporally consistent, gridded daily meteorological dataset for northwestern north america. Scientific Data, 6(1), 180299.