An inventory of historical climate data and climate projections for the Canadian North

Executive summary

Lead authors

Lead Authors: Emilia Diaconescu (CCCS/ECCC), Silvie Harder (CCCS/ECCC), Elaine Barrow (CCCS/ECCC), Jennifer Lukovich (University of Manitoba), Stephen Déry (University of Northern British Columbia), Lawrence Mudryk (CRD/ECCC), Rajesh Shrestha (WHERD/ECCC ), Alex Crawford (University of Manitoba), Paul Kushner (University of Toronto), Stephan Gruber (Carleton University)

Overview

A ‘Northern Climate Data Working Group’ was brought together by ECCC to advise on how to identify, evaluate, and select climate datasets for various regional and local applications in the Canadian North. This report constitutes the result of the Phase I of the Working Group.

Scope

The scope of this collaborative report is to identify, inventory, and characterize existing datasets for future development of climate products to support climate-change adaptation decision-making in the Canadian North. It constitutes an important first step in the complex task of developing climate services in this region.

Primary audience

The intended primary audience for this report comprises climate services organizations, academics, and consultants who require information on the suitability and availability of raw climate datasets to develop climate products and tools to fill the gap between the theoretical knowledge of processes and the practical application for adaptation.

Dataset categories and characteristics

Datasets for several variables, grouped in five categories (meteorology, snow,  hydrology, sea ice, and permafrost), were inventoried for the historical period and for future projections, with a more extended analysis for the former.

The characteristics of the selected variables are tabulated in the report with links to websites and descriptive documents, many of which are included as annexes. The annexes describe the metadata using a standard format, contain links to download the data, and additional details on the methodology used to develop the datasets.

The historical observational data and the modelled data are presented in distinct chapters.

Present understanding of datasets and the best practices for their use are also provided.

The following presents the key messages and recommendations grouped by category of variables.

Meteorological data and key messages

There are many datasets that provide long-term data for the four meteorological variables analyzed in this project. The report includes multiple annexes (the number indicated in brackets) for each of these variables: near-surface air temperature (15), total precipitation (27), surface humidity (9), and surface wind speed and direction (12).

Historical observational datasets

The best solution for describing local climate, trends and evolution of meteorological variables over a long historical time period is to use adjusted and homogenized data series from stations as provided in AHCCD datasets (see Annexes 7.1.5; 7.2.4 & 7.4.3). Those datasets take into account changes in instrumentation, location, and sampling, and correct spurious, non-climate-related shifts; the adjusted precipitation datasets account for a number of known errors in precipitation measurements. 

Despite their past use, datasets obtained by using gridding methods for regions of the Canadian North should be considered with caution where the station network is known to be sparse. Only regional applications in areas with a good coverage of stations (e.g., a small watershed that has several stations in each grid cell) and without important topographic variations could consider the use of a gridded-observed dataset (see discussions in Section 3.1.2 and Section 3.1.3). 

Reforecasts and reanalyses represent a valuable alternative to gridded-observed datasets for applications that require meteorological data over large regions without good station coverage or length of record, such as impact models, statistical downscaling and bias correction of climate models.

However, it is difficult to recommend one particular reanalysis for those applications as many of the new datasets have not yet been evaluated in a rigorous manner over the Canadian North. The results of continental-scale evaluations of these datasets could be biased by the much larger number of stations in regions outside the Canadian North. Several studies recommend a multi-dataset approach to compensate for the differences found in the various reanalyses. 

While merged or “hybrid” datasets, which combine the observations from multiple satellite platforms, constitute a new and potentially useful source of historical precipitation data (see Section 3.1.3),  there are no studies that evaluate them over the Canadian North.

Although specific applications require long-time series of sub-daily data, only a limited number of stations, reanalyses and climate models have them.

Climate projections

For climate projections, it is recommended to use multiple GCMs/ESMs/RCMs, or statistically downscaled GCMs, and multiple RCPs/SSPs in any given analysis to account for model and emission uncertainties.

Model selection depends on the application and the climate variable. Bias-corrected simulations exist for air temperature and precipitation, with most of these using a gridded observed dataset for reference for the historical period. A very limited number of analyses were done for climate projections of humidity and wind over Canada.  

Snow data and key messages

Three snow variables were considered in this report (snow depth, snow water equivalent (SWE) and snow cover) and 25 observation-based  datasets are described in the appendices.

Historical observational datasets 

While there are numerous datasets available to assess snow variables, there are strengths and limitations depending on the data type (station data, gridded data, satellite data or modelled data) and even among individual datasets within a given type. Selection of a specific snow dataset depends on the application and the spatial and temporal scale of interest.

Snow variables can vary significantly at the local scale because of interactions with vegetation, topography and wind. In situ or site-specific measurements can offer long-term records that reflect the local variability, but they are sparsely distributed in the Canadian North and may not capture the average characteristics of the surrounding snow conditions (see additional information provided in Section 3.2.1). 

Gridded historical products, both satellite-derived and reanalysis-based, offer spatially and temporally complete data, but contain many assumptions and/or modelled components and are rarely (if ever) evaluated specifically for their ability to accurately reproduce conditions in the Canadian North.

Research results suggest that at least for historical, gridded, hemispheric SWE datasets, averaging multiple products together can improve accuracy, but this is untested on smaller regional scales (e.g., the Canadian North) and may or may not apply to other snow variables.

Climate projections 

Projected changes of snow variables are typically obtained from CMIP5/6 (GCM) or CORDEX (RCM) multi-model ensembles for several emission scenarios (see additional information on the present skill of simulated snow in Section 4.2.3). 

Data products based on applying statistical downscaling directly to snow variables are generally not available. It is possible to obtain dynamically downscaled snow data by running high resolution models with downscaled/bias-corrected meteorological forcing but at present these are research products. 

Hydrological data and key messages

River discharge was summarized in the hydrology data category (see Section 3.2.2 for details on observations and Section 4.2.2.2 for details on modelled data) and 8 observation-based  datasets are described in the appendices.

Historical observational datasets 

Historical datasets on river discharge rely mainly on in situ measurements of water levels which are converted to discharge using a rating curve. Emerging remote sensing technology, and numerical models (e.g., hydrologic and hydraulic models) offer additional sources of data, filling gaps in ungauged areas.

Availability of hydrometric data across northern Canada expanded markedly in the 1960s but suffered declines in the 1990s.

Principal rivers of northern Canada are gauged by the Water Survey of Canada (sometimes by hydropower companies) but smaller rivers/streams often lack hydrometric stations, most notably in the Canadian Arctic Archipelago. Thus, remote sensing, reanalyses and model output complement the observed data, but their reliability is highly variable  in ungauged basins of northern Canada and they are generally limited to recent decades at most. Many gauging stations are not monitored in winter, or when ice is present.

The recommendation for the historical datasets is to give preference to in situ river discharge  from hydrometric stations, where and when available, given they generally have extended and homogeneous records.

When discharge data are insufficient or simply unavailable, other sources including experimental watersheds maintained by academic institutions and government agencies, and reconstructed streamflow data through remote sensing or models may be employed.

Climate projections

Basin-scale hydrologic models, global/regional hydrology models and land surface models can be used to provide GCM-driven streamflow simulations over historical and future periods in the Canadian North.

Unlike the basin-scale hydrologic models, the regional/global models are generally not calibrated to reproduce the basin-specific streamflow, which can lead to considerable uncertainty.

It is a common practice to incorporate multiple emissions scenarios, an ensemble of multiple GCMs or RCMs and one or several basin-scale hydrological models in projecting hydrologic impacts of climate change.

The simulations from the basin-scale hydrologic models are generally considered suitable for detailed analyses (e.g., seasonal water availability, extreme events), while those from the global/regional hydrology models and RCMs can be used to obtain large-scale general information (e.g., the direction of changes in streamflow over large areas).

Sea-ice data and key messages

Three sea ice variables were analyzed in this report and the recommended datasets for the Canadian North are described in sections 3.3.2 (sea ice concentration), 3.3.3 (sea ice thickness) and 3.3.4 (sea ice drift).

Historical observational datasets

The sea ice charts from the Canadian Ice Service Digital Archive provide a local and regional characterization of sea ice types; however, they can be subject to differences in interpretation, resulting in features that depart from actual local conditions.

Gridded sea ice concentration datasets capture regional and hemispheric-scale changes in sea ice, but they are subject to uncertainties associated with cloud cover, snow on sea ice, melt ponds and coarse resolution, which hinder accurate representation of landfast ice, ice edge location, and marginal ice zone extent. 

In-situ sea ice thickness observations provide a local characterization of changes in sea ice volume, while satellite-derived sea ice thickness provides continuous spatial and temporal coverage albeit for a shorter historical record than for sea ice concentrations and drift.

Consistent in-situ sea ice thickness measurements in the Canadian Arctic exist only for Cambridge Bay, Resolute, Eureka, and Alert from 1950 to present. Recent studies have shown that CryoSat-2 satellite measurements provide a reliable estimate of thicknesses ranging from 0.5 to 4 m, while a combined CrysoSat-2 and Soil Moisture and Ocean Salinity product provides a reasonable estimate of thin ice; however, they are available for a short period of time only. 

For sea ice drift, IABP provides small-scale, high-frequency measurements of sea ice trajectories from drifting buoys that are not uniformly spatially or temporally distributed.

NSIDC sea ice drift provides continuous spatial and temporal coverage of sea ice drift at a spatial resolution of 25 km from 1979, and OSISAF sea ice drift at a spatial resolution of 62.5 km from 2009, with summertime drift available following 2017. However, persistent features due to the merging of buoy and satellite data in the NSIDC dataset, and coarse spatial resolution in the OSISAF dataset hinder evaluation of sea ice motion gradients, or deformation. 

For applications, it is recommended that users consider and account for differences in algorithms and methods used to combine sources into long-term sea ice datasets. For long-term and large-scale scientific research, the passive microwave record of sea ice (using the NOAA/NSIDC Climate Data Record version) is recommended because it covers a broader area and yields all three key sea ice variables from one product.

For applications in the Canadian North, the sea ice charts have been shown to provide more accurate estimates of sea ice concentration in Canadian waters compared to satellite passive microwave estimates. It is recommended that local sea ice information be supplemented with in-situ measurements that can be found through the Polar Data Catalogue (PDC),  SIKU, and Smart-ICE initiatives (Section 3.3.5). Indigenous knowledge should inform and address specific questions and applications pertaining to Indigenous communities.

Climate projections 

For climate projections of sea-ice variables, the present practice is to use the CMIP5/6 multi-model ensembles for several emission scenarios (see Section 4.2.3). There are insufficient downscaled products and no bias-corrected products for sea ice, and the narrow channels in Canadian Arctic waters are not well represented by the coarse spatial-resolution GCMs. While uncertainty in model projections is higher for the Canadian Arctic Archipelago than for the pan-Arctic, the CMIP5/6 multi-model ensemble still provides a quantitative basis for projecting future sea ice conditions.

Permafrost Data and key messages

Permafrost was analyzed in this report in section 3.4. Datasets included in this section are ground temperature (3.4.2), subsurface ice content (3.4.3) and permafrost extent (3.4.4). Permafrost change is described in terms of landform inventories (3.4.5) and ground subsidence and active-layer thickness (3.4.6).  See section 4.2.4 for details on modelled data.

Historical observational datasets 

Ground temperature is an important and challenging metric of permafrost change as permafrost is defined by temperature, and interest in permafrost is largely due to the observed thawing and its impacts. Ground temperature datasets are often not widely shared between researchers or different groups, or the data are not in a standardized form and are only available for certain sites and studies. PermafrostNet is working to improve data standards and interoperability (see section 3.4.2).

Interpretation of ground temperature conditions can be difficult as they vary widely even on local scales. This also means that interpretation of gridded data can be difficult on a fine scale. Reliable ground temperature data are useful for engineering design, the quantification of change and for the testing of climate models. 

The Global Terrestrial Network for Permafrost (GTN‐P) is an international programme concerned with monitoring permafrost parameters and Permafrost Ground Temperature for the Northern Hemisphere, v3.0 provides modelled data products. There are also many region-specific borehole and ground temperature databases available across northern Canada. 

Ground ice formation and melt can have a big impact on topography, local hydrology and infrastructure which means that the quantity, location and type of ground ice are important when considering how an area with ice-rich permafrost may respond to climate change (see section 3.4.3).

There are several ground-ice maps of the circum-Arctic and of Canada where ground ice conditions are represented by either modelled or borehole data. Both types of data have strengths and limitations: Borehole data are valuable for the characterization of local ice conditions but the data can vary depending on the drilling campaign.

Modelled data can provide estimates of ground ice content over large areas but there is a general lack of ground ice information for calibration and validation of models. This means that most models and maps created by modelled data can provide only a rough estimate of average ground ice conditions within the map units.

Permafrost extent, broken into zones such as continuous, discontinuous or extensive discontinuous permafrost, expresses the areal proportion of ground underlain by permafrost. Permafrost extent cannot be directly observed and data products representing this variable are based on models that rely usually on air and ground temperature observations.

Quantifying permafrost extent and changes in permafrost extent often requires extreme simplification and can limit the interpretability of results (see section 3.4.4).

There are several historical datasets available for permafrost extent, generated either by manual delineation as polygons or as gridded models that rely on computer simulations. The Circum-Arctic Map of Permafrost and Ground-Ice Conditions is the most widely used spatial depiction of permafrost.

Permafrost landform inventories can contribute to anticipating future trajectories of change through insight into past, subsurface processes through their surface expression as landforms.

While landform inventories exist for different areas of the Canadian Arctic, they are not always systematically collected and are often tied to specific research questions. There is a need to define landforms objectively and reproducibly (see section 3.4.5).

There are several data sources and mapping initiatives across northern Canada that are working to create thermokarst and permafrost feature inventory maps through manual delineation and identification of landforms or by gridded mapping. 

The Circumpolar Active Layer Monitoring (CALM) program is the primary programme for long-term active-layer measurements (the layer above permafrost that freezes and thaws annually). Ground subsidence occurs with increases in active-layer thickness where the sediments are ice-rich. Data on active layer thickness and ground subsidence in Canada are currently rare and not organized into a central, accessible repository, although this information would improve knowledge of landscape changes associated with permafrost thaw in ice-rich terrain (section 3.4.6).

Climate projections 

Even though variables related to ground temperature are available from climate projections, their application for informing decision-making related to future permafrost thaw is limited See section 4.2.4 for a discussion on permafrost and climate models.

The report of the Phase I of the Working Group includes the inventory of datsets for the five categories of variables and the online version can be accessed on the Northern Climate Data Report and Inventory (NCDRI) website.

Recommended Citation for the report:

Diaconescu, E.P., P. Kushner, J. Lukovich, A. Crawford, E. Barrow, L. Mudryk, M. Braun, R. Shrestha, S. Gruber, S. Déry, S. Howell, and L. Matthews, 2023: An inventory of historical climate data and climate projections for the Canadian North; Government of Canada, Gatineau, QC, 698p

Contact us

If you would like to obtain the PDF version of the Phase I report of the Working Group, it is available through the Canadian Centre for Climate Services (CCCS). Please reach out to CCCS using the following email address:

Center: Canadian Centre for Climate Services (CCCS)/ECCC

Email: ccsc-cccs@ec.gc.ca