1) If you are solely interested in understanding the past conditions that were recorded by meteorological instruments, and if your location of interest is near a weather station, then you would use the station data observations available on ClimateData.ca.
2) However, weather station observations can be affected by changes in instruments and the surroundings of the meteorological station where they were recorded. If you are interested in understanding the trends in past conditions at specific locations, you should use a dataset called Adjusted and Homogenized Canadian Climate Data (AHCCD). This dataset has been developed to account for changes at weather stations, including changes in observing technology, which are not due to changing weather conditions. For example, new instruments may be installed, trees may have grown up around the site, or the station may have been moved slightly for some reason. All of these can affect the recorded observations. Scientists have spent considerable effort checking weather records and making sure that these changes are accounted for. You can find the AHCCD on the Canadian Centre for Climate Services’ website.
3) If you are interested in understanding past conditions in areas away from stations, then the gridded, ANUSPLIN data could be used. It may be possible to get access to ANUSPLIN data for your location of interest. You can contact the Support Desk for dataset support and guidance.
4) If you want to compare future projections with past conditions, then you should use the modelled historical data. Climate models are mathematical representations of the real world, but their simulations of historical climate are not exactly the same as actual observed climate, Climate models simulate average values over areas that can be 10,000 square km in size or larger, while weather stations provide observed values at specific locations. Also, because our understanding of the climate system is still not complete, climate models may show biases – systematic differences from observed values. There are mathematical and statistical methods that can reduce this bias, and these have been applied to the climate model datasets available on ClimateData.ca. Although some biases will inevitably remain, these datasets do exhibit similar average values and variability when compared with observed conditions over the historical period, but their time sequencing of individual weather variations is different. This does not mean that the model data is incorrect – the time sequencing is expected to be different. The climate models simulate weather with statistics that are similar to those of the observed weather, but they are not designed to forecast the exact sequence of weather events that has been observed over the historical observing period. For all of these reasons, it is best to compare the future projections from the climate models with the historical simulations from the same models rather than with historical observations.