This article provides a high-level perspective on existing weather files for Canada developed from historical data, and on future weather files that offer information about building design conditions in a changing climate. The Pacific Climate Impacts Consortium (PCIC) and the National Research Council of Canada (NRC) have each developed a set of future weather files using different but conceptually convergent methods.
The Canadian Weather Year for Energy Calculation (CWEC) dataset provides historical weather files commonly used by building designers, engineers, and energy modelers for evaluating the performance of building designs. The CWEC weather files are developed by the Engineering Climate Services Unit of the Meteorological Service of Canada (MSC) that are also responsible for providing data for the National Building Code of Canada.
The CWEC dataset uses the Typical Meteorological Year (TMY) methodology, which involves selecting the twelve most ‘typical’ months from a multi-year period of record (15, 20 or 30 years of continuous data), to create a single artificial year that represents median weather conditions. Table1 summarizes the actual year from which the hourly historical data has been selected for each month to create the CWEC weather file for Toronto international airport. For example, January is copied directly from the historical January data that has the closest match to the period of record’s average condition for all January’s. This shows that the contributing year to each month can change between each CWEC iteration.
| CWEC | CWEC2016 | CWEC2020 | |
| Period of Record | 1959 to 1989 | 1998 to 2014 | 1998 to 2017 |
| Months | |||
| Jan | 1969 | 1999 | 2007 |
| Feb | 1965 | 2004 | 2004 |
| Mar | 1964 | 2006 | 2002 |
| Apr | 1964 | 2009 | 2009 |
| May | 1964 | 2006 | 2006 |
| Jun | 1970 | 2001 | 2017 |
| Jul | 1981 | 2013 | 1998 |
| Aug | 1989 | 2011 | 2013 |
| Sep | 1978 | 2003 | 1998 |
| Oct | 1969 | 2010 | 2014 |
| Nov | 1983 | 2000 | 2017 |
| Dec | 1961 | 2003 | 2004 |
In the statistical selection of ‘typical’ months, variables are assigned weighting factors based on their relative importance and sensitivity in building performance simulation. As such, the weighting factors assigned to each variable play a critical role in the selection of the typical months for inclusion in the weather files. Higher weighting factors applied to temperature and solar radiation, for example, can result in the selection of typical months more representative for those variables, while potentially being less representative for values of other variables stored in the file.
While the TMY methodology aims to produce a single year of hourly climate data that is representative of a given location, it does not necessarily capture the climatology of the entire period of record from which the data is chosen – for two reasons. First, the median is not necessarily the mean; second, the weighting of multiple variables can cause the final result to differ from the median for individual variables.
CWEC dataset is currently the best-in-class weather file-based assessments of historical climate conditions for building simulation in Canada. The CWEC weather files are a valuable tool for approximating typical conditions for a given location, but they are an artificial year, and it is important to understand the selection process, weighting factors, and the period of record, as not only these effects building simulation results using historical weather files but can also influence future weather files that are developed. To support climate resilient building design, it is suggested to use both historical and future weather files in building simulation. The following sections describe how two future weather file datasets are developed for Canada.
To create suitable climate data for building simulation applications, projections from one or several global climate models (GCMs) are needed. However, the outputs from GCMs are often provided at large spatial scales (i.e., global, continental, or 100 km scale) and daily or longer temporal resolutions (A limited number of GCMs provide sub-daily simulations). This is too coarse for direct use in building simulations that require, at the minimum, hourly data for a local area. As a result, one of two downscaling techniques (See Box 2.2) are used to adjust the GCM outputs and obtain finer resolution scenarios of climate change from the original coarser resolution GCM information.
In addition, the GCM output can contain some biases (errors) caused by a range of factors, such as simplified representation of some physical processes like cloud formation and their interaction in climate models or incomplete understanding of the global climate system. Therefore, it is also important to bias adjust climate models. The quality of the adjustment relies on the availability of high-quality reference datasets.
The first technique is statistical downscaling. Statistical downscaling establishes statistical relationships between historical observed climate data and the output of the climate model for the same location and time-period. Once this relationship has been determined, future projections from the climate model can be used to determine the future climate values at the local scale. A common statistical downscaling method employed in Canada is known as the BCAAQv2 methodology, used in producing the Canadian Downscaled Climate Scenarios (CanDCS).
The second downscaling technique is dynamical downscaling, where the output from a GCM is used to drive a Regional Climate Model (RCM). RCMs, like GCMs, are mathematical representations of the physical processes and interactions between all parts of the climate system. An RCM is used to enhance the spatial resolution of the selected region (i.e., North America, Europe) and thus provides a more accurate representation of the underlying topography and land/water boundaries. The spatial resolution of RCMs is generally between 10 and 50 km. This downscaling technique is typically orders of magnitude more computationally expensive than statistical downscaling.
The future weather files developed by PCIC and NRC are different in terms of:
To obtain suitable climate information for building simulation from GCM output, PCIC statistically downscaled and bias adjusted daily climate projections from an ensemble of 10 climate models. For dry-bulb temperature values, PCIC used the BCCAQv2 methodology to downscale climate projections from GCMs to a gridded resolution of roughly 10 km x 6 km. For dew-point temperature, relative humidity and surface pressure, climate projections from the GCMs were only interpolated to the BCCAQv2 resolution.
Next, PCIC incorporated these statistically downscaled and bias adjusted projections into weather files using the “morphing technique”1. The morphing technique applies three transformation functions (shift, stretch, and their combination) to adjust the hourly values of any given variable stored in a historical weather file. For this reason, PCIC’s future weather files are often referred to as future-shifted weather files. The process of applying the projected climate change to the historical CWEC weather files highlights the importance of the period of record as any differences in this period will propagate to the future weather files.
Adjustments to hourly values in PCIC future-shifted weather files are applied to dry-bulb and dew point temperatures, relative humidity, and surface pressure. All other variables stored in the historical weather file remain unadjusted. Therefore, the hourly values of solar radiation, wind speed/direction, cloud cover, etc., in the historical weather file – CWEC2016, period of 1998 to 2014 – that is used as the baseline period of record and the three generated future weather files are identical. The results are three, 30-year future-shifted weather files, using RCP8.5, for the following time-periods: 2020s (2011 to 2040), 2050s (2041 to 2070), 2080s (2071 to 2100).
Future weather files developed by NRC are obtained directly from dynamically downscaled and bias adjusted climate projections using the second-generation Canadian Earth System Model, CanESM2 (A coarse resolution global model). This work was built upon the framework developed by the Climate-Resilient Buildings and Core Public Infrastructure (CRBCPI) project, where the CanESM2 large ensemble – consisting of 50 simulation runs of the CanESM2 model with different initial conditions – was dynamically downscaled using the Canadian Regional Climate Model Version 4 (CanRCM4). NRC used a subset of the CanRCM4 large ensemble simulations comprising of 15 runs to acquire the climate projections. Some climate variables were not directly available from the CanRCM4 large ensemble at hourly scale and hence were estimated: rainfall (from precipitation), snow-cover (from snow-depth), as well as direct horizontal, direct normal, and diffused horizontal solar radiation (using established methods from literature).
NRC used the following steps to generate their future weather files:
More detailed information on the future weather files method can also be accessed in PCIC and NRC’s technical documents. For guidance on using these future weather files, see the article “Guidance on Using Future Climate Data for Building Performance Simulation”.