Statistically Downscaled Climate Scenarios

PCIC offers statistically downscaled daily Canada-wide climate scenarios, at a gridded resolution of 300 arc-seconds (0.0833 degrees, or roughly 10 km) for the simulated period of 1950-2100. The variables available include minimum temperature, maximum temperature, and precipitation. Users may access the scenarios using an interactive map interface that allows users to zoom, pan and select their region of interest using a rectangular-selection tool.

CLIMATE MODELS AND SCENARIOS

A. Coupled Model Intercomparison Project Phase 5 (CMIP5)

Downscaled scenarios were constructed from 27 Global Climate Models (GCMs) and 3 Representative Concentration Pathways (RCPs) (van Vuuren et al., 2011) from CMIP5 (Taylor et al., 2012) using the BCCAQv2 downscaling method described below. Scenarios can be selected from any combination of models, RCPs, and time period.

This data set is referred to as Canadian Downscaled Climate Scenarios – Univariate (CMIP5), or CanDCS-U5 for short.

Access and Download CanDCS-U5

B. Coupled Model Intercomparison Project Phase 6 (CMIP6)

Downscaled scenarios were constructed from 26 GCMs and 3 Shared Socioeconomic Pathways (SSPs) (Riahi et al., 2017) from CMIP6 (Eyring et al., 2016) using the two distinct downscaling methods described below. Scenarios can be selected from any combination of models, SSPs, and time period.

STATISTICAL DOWNSCALING METHODS

Data from each climate model is downscaled to a finer resolution using one or more statistical methods and a gridded “target” observation-based dataset, which constitutes a reconstruction of the actual historical climate over Canada.

Note that for the historical period used to calibrate the downscaling methods (1951-2005 for CMIP5, 1951-2012 for CMIP6) , statistical properties of the downscaled results will, by design, tend to match those of the gridded observational dataset. The day-by-day, month-by-month, year-by-year, etc. sequencing of values, however, will not correspond to observations, since climate models solve a “boundary value problem” and are not constrained to reproduce the timing of natural climate variability (e.g., El Niño-Southern Oscillation) in the observational record.

A. N-Dimensional Multivariate Bias Correction (MBCn)

The MBCn technique (Cannon 2018; see also Francois et al., 2020) adapts the statistical characteristics of a multivariate reference distribution (i.e., gridded observations) to the multivariate distribution of climate model variables. The method begins with the univariate adjustment of each climate variable, using the same QDM procedure as in BCCAQv2. Next, the dependence structure between variables is adjusted using a multi-step, iterative process based on an adaptation of an image processing algorithm used to transfer color information. The iterative process converges when the multivariate distributions of the reference observations and historical climate simulations agree to within a specified tolerance; by design, adjusted univariate distributions match those of QDM. Furthermore, MBCn reflects changes in the dependence structure amongst variables in the future climate simulations.

This data set is referred to as Canadian Downscaled Climate Scenarios – Multivariate (CMIP6), or CanDCS-M6 for short.

Access and Download CanDCS-M6

B. Bias Correction/Constructed Analogues with Quantile Mapping Reordering (BCCAQv2)

BCCAQv2 is a hybrid method developed at PCIC that combines results from Bias Corrected Constructed Analogs (BCCA; Maurer et al. 2010) and Quantile Delta Mapping (QDM; Cannon et al. 2015). BCCA uses spatial aggregation from a linear combination of historical analogues for daily large-scale fields. QDM applies a form of quantile mapping where relative changes in GCM quantiles are preserved to avoid inflationary effects that can occur with standard quantile mapping. BCCAQv2 is an updated version of BCCAQ (version 1), which employed standard quantile mapping.

This data set is referred to as Canadian Downscaled Climate Scenarios – Univariate (CMIP6), or CanDCS-U6 for short.

Access and Download CanDCS-U6

C. Other Downscaling Methods 

Other methods used at PCIC for previous version of our downscaled products include the following. 

BCCAQv1: The original version of BCCAQ differs from BCCAQv2 (see above) in that it uses QMAP for the quantile mapping step (Gudmundsson et al., 2012). 

Bias-corrected Spatial Disaggregation (BCSD): BCSD (Wood et al., 2004; Maurer et al., 2008) was used with the following modifications: the incorporation of monthly minimum and maximum temperature instead of monthly mean temperature, as suggested by Bürger et al., (2012) and bias correction using detrended quantile mapping with delta method extrapolation following Bürger et al., (2013). For past validation and analysis of the BCSD downscaling algorithm for British Columbia, see Werner (2011), Bürger et al. (2012, 2013) and Werner and Cannon (2016).

DATASET VERSIONS

Versions of the CanDCS datasets are listed in Table 1 below and the most recent include the files currently available for download. Updated versions include minor revisions to aspects of the datasets and are described in Table 1. The "Variables" column indicates which of the CanDCS variables have been updated in the revised versions (PR: precipitation, TX: maximum temperature, TN: minimum temperature).

Table 1. File versions.

MODEL SUBSETS (CMIP6)

Analysis and storage of the full climate model ensemble is often not feasible. For the CMIP5 ensemble, a suggested subset of 12 models was provided, based on a method designed to capture 90% of the range of projected changes in temperature and precipitation in all seasons for the Climdex suite of extreme indices of extremes under RCP4.5 (the KKZ method, see Cannon, 2015). For CMIP6, this method was improved in two main respects. First, the full set of 26 GCMs was screened to account for models that have a significant “shared history” of development, which produces redundancy in model outputs (Brunner et al. 2020). Second, the issue of scenario dependence was addressed by use of a cross-scenario selection strategy, specifically by applying the KKZ method to future climate index changes for all SSPs at a specified level of global warming, namely 2 ℃. Additional improvements included using the downscaled (instead of raw) GCM results to compute indices and the removal of redundant indices from the procedure.

For CMIP6, recommended subsets of models are provided in the Tables below for Canada along with a number of its sub-regions shown in the accompanying map The national subset comprises 12 GCMs and is identical for both CanDCS-U6 and CanDCS-M6. Users interested in climate projections for a single region are advised to use the corresponding regional subset listed in the Tables, while those studying than one region should use the Canada-wide subsets. The number of GCMs in the regional subsets ranges from 8 to 11, depending on the downscaling method and region. As for the national subsets, there is a high degree of overlap between the downscaled GCMs from the two methods.

Further details about this selection method can be found in the project report, Downscaled CMIP6 Climate Model Subset Selection (PCIC, 2023).

Table 2. Cross-scenario, KKZ-selected GCM subsets for Canada and each of its subregions, from the reduced set of CanDCS-U6 downscaled CMIP6 scenarios. Models in grey shaded cells differ between the CanDCS-U6 and CanDCS-M6 subsets. Note that in the tables, the numeric ordering should not be interpreted as a quality-based performance ranking. Rather, the first-ranked GCM has a mean change over all climate indices which is closest to the change obtained from the entire CanDCS-M6 ensemble, while subsequently ranked models incrementally sample the range in future changes found in the full ensemble. 

Table 3. Cross-scenario, KKZ-selected GCM subsets for Canada and each of its subregions, from the CanDCS-M6 downscaled CMIP6 scenarios. Models in grey shaded cells differ between the CanDCS-U6 and CanDCS-M6 subsets.

Figure 1: Regions of Canada used for downscaled GCM subset selection.

MODEL SUBSETS (CMIP5)

The 12 GCMs listed in the Table below capture 90% of the range of projected changes in temperature and precipitation in all seasons for a suite of indices of extremes under RCP4.5 (Cannon, 2015). Recommended subsets of models are provided for a number of geographic sub-regions of North America, known as Giorgi regions (Giorgi and Francisco, 2000), shown in the map below the Table. Note, however, that data are only available for the parts of the Georgi regions that lie within Canada.  If fewer than 12 GCM runs are desired, they should be chosen following the order listed for each sub-region in the Table. Note that only 9 of the 12 GCM runs are available for RCP2.6.

The 12 GCM runs listed under the WNA region are used in PCIC’s Plan2Adapt and PCIC’s Climate Explorer online tools (PCEX).

Table 4. Model Ensembles and Giorgi Regions.

Figure 2: Giorgi regions that intersect with Canada: Alaska (ALA), Western North America (WNA), Central North America (CNA), Greenland (GRL), Eastern North America (ENA) and Central America (CAM).

DOWNSCALING TARGET DATASETS

A. CMIP6 (CanDCS-M6)

The daily, gridded observational dataset used to calibrate MBCn, referred to as PCIC-Blend, is based on three existing data sets. Two of these are recently upgraded versions of the NRCANmet dataset used to create CanDCS-U5 and CanDCS-U6: NRCANmet-Adjusted Precipitation, which spans Canada (MacDonald et al. 2021), and NRCANmetV2 Temperature, spanning North America (MacDonald et al. 2020). The third dataset, PNWNAmet, covering Western Canada and the Pacific Northwest, is available on the PCIC Data Portal under Daily Gridded Meteorological Datasets. While the updated NRCANmet data sets display notable improvements over the earlier version over central and eastern Canada, their performance is inferior to PNWNAmet over western Canada when compared to highquality station observations. Hence, PNWNAmet values were blended with the NRCANmet values over a small overlap region stretching from southwest Alberta to the northeast Yukon, using NRCANmetV2 for temperature and NRCANmet-Adjusted for precipitation. See the Daily Gridded Meteorological Datasets page for further details on the construction of PCIC-Blend.

B. CMIP5 (CanDCS-U5) AND CMIP6 (CanDCS-U6) 

The gridded observational dataset used to calibrate BCCAQv2, referred to as NRCANmet, was produced by Natural Resources Canada (NRCan) and is available at 300 arc second spatial resolution (1/12° grids, ~10 km) over Canada. Daily minimum and maximum temperature, and precipitation amounts for the period 1950-2012 were produced by Hopkinson et al. (2011) and McKenney et al. (2011) on behalf of the Canadian Forest Service (CFS), NRCan.  Gridding was accomplished with the Australian National University Spline (ANUSPLIN) implementation of the trivariate thin plate splines interpolation method (Hutchinson et al., 2009) with latitude, longitude and elevation as predictors. This dataset is also available via the PCIC Daily Gridded Meteorological Datasets page. Note that gridded values may differ from climate stations and biases may be present at high elevations or in areas with low station density (Eum et al., 2014).

ACKNOWLEDGEMENTS

We thank the Landscape Analysis and Applications section of the Canadian Forest Service, Natural Resources Canada for developing and making available the Canada-wide historical daily gridded climate datasets used as the downscaling targets. We acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP5 and CMIP6, and we thank the climate modeling groups for producing and making available their GCM output. For CMIP the U.S. Department of Energy’s Program for Climate Model Diagnosis and Intercomparison provides coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Science Portals. PCIC gratefully acknowledges support from Environment and Climate Change Canada for the development of the statistically downscaled GCM scenarios that are distributed from this data page.

DATA CITATIONS

When referring to the Statistically Downscaled Climate Scenarios produced by PCIC, whether retrieved from this website or found otherwise, the source must be clearly stated. For the most recently released data, the following citation is recommended:

Citation for CanDCS-U5:
"Pacific Climate Impacts Consortium, University of Victoria, (Feb. 2019). Statistically Downscaled Climate Scenarios. Downloaded from https://data.pacificclimate.org/portal/downscaled_gcms/map/ on <date>. Method: BCCAQ v2.” (Please also specify GCMs, RCPs, domain of download, and downloaded variables. Citation of relevant references, as provided below, is also recommended where appropriate.)

Citation for CanDCS-U6:
"Pacific Climate Impacts Consortium, University of Victoria, (Dec. 2021). Statistically Downscaled Climate Scenarios. Downloaded from https://data.pacificclimate.org/portal/downscaled_cmip6/map/ on <date>. Method: BCCAQv2.” (Please also specify GCMs, SSPs, domain of download, and downloaded variables. Citation of relevant references, as provided below, is also recommended where appropriate.)

Citation for CanDCS-M6:
"Pacific Climate Impacts Consortium, University of Victoria, (July 2023). Statistically Downscaled Climate Scenarios. Downloaded from https://data.pacificclimate.org/portal/downscaled_cmip6/map/ on <date>. Method: MBCn.” (Please also specify GCMs, SSPs, domain of download, and downloaded variables. Citation of relevant references, as provided below, is also recommended where appropriate.)

TERMS OF USE

In addition to PCIC's terms of use, the data for each individual data set is subject to the terms of use of each source organization. For further details, please refer to:

• The terms of use for CMIP5
• The terms of use for CMIP6

NO WARRANTY

This data product is provided by the Pacific Climate Impacts Consortium with an open license on an “AS IS” basis without any warranty or representation, express or implied, as to its accuracy or completeness. Any reliance you place upon the information contained here is your sole responsibility and strictly at your own risk. In no event will the Pacific Climate Impacts Consortium be liable for any loss or damage whatsoever, including without limitation, indirect or consequential loss or damage, arising from reliance upon the data or derived information.

References:

Brunner, L., Pendergrass, A. G., Lehner, F., Merrifield, A. L., Lorenz, R., & Knutti, R., 2020: Reduced global warming from CMIP6 projections when weighting models by performance and independence. Earth System Dynamics, 11, 4, 995-1012, doi:10.5194/esd-11-995-2020. 

Bürger, G., T.Q. Murdock, A.T. Werner, S.R. Sobie, and A.J. Cannon, 2012: Downscaling extremes - an intercomparison of multiple statistical methods for present climate. Journal of Climate, 25, 4366–4388. doi:10.1175/JCLI-D-11-00408.1.

Bürger, G., S.R. Sobie, A.J. Cannon, A.T. Werner, and T.Q. Murdock, 2013: Downscaling extremes - an intercomparison of multiple methods for future climate. Journal of Climate, 26, 3429-3449. doi:10.1175/JCLI-D-12-00249.1.

Cannon, A. J., 2018: Multivariate quantile mapping bias correction: an N-dimensional probability density function transform for climate model simulations of multiple variables. Climate Dynamics, 50, 31-49, doi:10.1007/s00382-017-3580-6.

Cannon, A.J., 2015: Selecting GCM scenarios that span the range of changes in a multimodel ensemble: application to CMIP5 climate extremes indices. Journal of Climate, 28, 3, 1260-1267. doi:10.1175/JCLI-D-14-00636.1

Cannon, A.J., S.R. Sobie, and T.Q. Murdock, 2015: Precipitation by Quantile Mapping: How Well Do Methods Preserve Changes in Quantiles and Extremes? Journal of Climate, 28(17), 6938-6959, doi:10.1175/JCLI-D-14-00754.1.

Eum, H.-I., Y. Dibike, T. Prowse and B. Bonsal, 2014: Inter-comparison of high-resolution gridded climate data sets and their implication on hydrological model simulation over the Athabasca Watershed, Canada. Hydrol. Process., 28, 4250–4271. doi: 10.1002/hyp.10236.

Eyring, V., S. Bony, G.A. Meehl, C. Senior, B. Stevens, R.J. Stouffer and K.E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geoscientific Model Development, 9(5), 1937-1958.

Francois, B. et al., 2020: Multivariate bias corrections of climate simulations: which benefits for which losses? Earth System Dynamics, 11, 537–562, doi:10.5194/esd-11-537-2020. 

Giorgi, F. and Francisco, R., 2000: Evaluating uncertainties in the prediction of regional climate change. Geophysical Research Letters, 27(9), 1295-1298, doi:10.1029/1999GL011016.

Gudmundsson, L., J. B. Bremnes, J. E. Haugen, and T. Engen-Skaugen, 2012: Technical Note: Downscaling RCM precipitation to the station scale using statistical transformations - a comparison of methods. Hydrology and Earth System Sciences, 16, 3383–3390, doi:10.5194/hess-16-3383-2012.

Hiebert, J., A. Cannon, A. Schoeneberg, S. Sobie, and T. Murdock, 2018: ClimDown: Climate Downscaling in R. The Journal of Open Source Software, 3(22), 360, doi:10.21105/joss.00360.

Hopkinson, R.F., D.W. McKenney, E.J. Milewska, M.F. Hutchinson, P. Papadopol, and L.A. Vincent, 2011: Impact of Aligning Climatological Day on Gridding Daily Maximum–Minimum Temperature and Precipitation over Canada. Journal of Applied Meteorology and Climatology, 50, 1654–1665. doi:10.1175/2011JAMC2684.1.

Hunter, R. D. and R. K. Meentemeyer, 2005: Climatologically Aided Mapping of Daily Precipitation and Temperature. Journal of Applied Meteorology, 44, 1501–1510, doi:10.1175/JAM2295.1.

MacDonald, H. et al., 2020: North American historical monthly spatial climate dataset, 1901–2016. Scientific Data, 7, 411. doi:10.1038/s41597-020-00737-2.

MacDonald H., D.W. McKenney, X.L. Wang, J. Pedlar, P. Papadopol, K. Lawrence, Y. Feng and M.F. Hutchinson, 2021: Spatial Models of Adjusted Precipitation for Canada at Varying Time Scales. Journal of Applied Meteorology and Climatology, 60, 3, 291–304, doi:10.1175/JAMC-D-20-0041.1.

Maurer, E.P., and H.G. Hidalgo, 2008: Utility of daily vs. monthly large-scale climate data: an intercomparison of two statistical downscaling methods. Hydrology and Earth System Sciences, 12, 2, 551-563. doi:10.5194/hess-12-551-2008.

Maurer, E., H. Hidalgo, T. Das, M. Dettinger, and D. Cayan, 2010: The utility of daily large-scale climate data in the assessment of climate change impacts on daily streamflow in California. Hydrology and Earth System Sciences, 14, 6, 1125–1138, doi:10.5194/hess-14-1125-2010.

McKenney, D.W., M.F. Hutchinson, P. Papadopol, K. Lawrence, J. Pedlar, K. Campbell, E. Milewska, R. Hopkinson, D. Price, and T. Owen, 2011: Customized spatial climate models for North America. Bulletin of the American Meteorological Society, 92, 12, 1611-1622. doi:10.1175/2011BAMS3132.1.

Riahi, K. et al. 2017:  The Shared Socioeconomic Pathways and their energy, land use, and greenhouse gas emissions implications: An overview. Global Environmental Change, 42, 153-168, doi:10.1016/j.gloenvcha.2016.05.009.

Sobie, S.R., and T.Q. Murdock, 2017: High-Resolution Statistical Downscaling in Southwestern British Columbia. Journal of Applied Meteorology and Climatology, 56(6), 1625–1641, doi:10.1175/JAMC-D-16-0287.1.

Taylor, K.E., R.J. Stouffer, and G.A. Meehl, 2012: An Overview of CMIP5 and the Experiment Design. Bulletin of the American Meteorological Society, 93, 485–498. doi: 10.1175/BAMS-D-11-00094.1.

van Vuuren, D.P., et al., 2011: The Representative Concentration Pathways: An Overview. Climatic Change, 109 (1-2), 5-31, doi:10.1007/s10584-011-0148-z.

Werner, A.T., 2011: BCSD downscaled transient climate projections for eight select GCMs over British Columbia, Canada. Pacific Climate Impacts Consortium, University of Victoria, Victoria, BC, 63 pp.

Werner, A. T. and A. J. Cannon, A. J., 2016: Hydrologic extremes – an intercomparison of multiple gridded statistical downscaling methods, Hydrololgy and Earth System Sciences, 20, 1483-1508, doi:10.5194/hess-20-1483-2016.

Wood, A.W., L.R Leung, V. Sridhar, and D.P. Lettenmaier, 2004: Hydrologic implications of dynamical and statistical approaches to downscaling climate model outputs. Climatic Change, 62, 189–216, doi:10.1023/B:CLIM.0000013685.99609.9e.