About BCCAQv2

Method overview

BCCAQV2 (Bias Correction/ Constructed Analogues with Quantile mapping reordering, Version 2) is a bias correction and statistical downscaling method for Global Climate Model (GCM) data, developed at PCIC (the Pacific Climate Impacts Consortium). Downscaling increases the spatial resolution of climate model data – transforming information at scales of hundreds of kilometres down to tens of kilometres – and thus making the data much more relevant at the local scale. Bias correction, or adjustment, reduces the difference (bias) between the modelled and observed data. BCCAQv2 has been widely used for downscaling daily climate model projections of temperature and precipitation. It is a univariate method which means that it downscales one variable at a time.

The CMIP5 & CMIP6 (Coupled Model Intercomparison Project) datasets which are downscaled using this method are the most widely-used downscaled climate data product in Canada.

Code implementing the BCCAQv2 method is also available via the R package “ClimDown”

How the method works

BCCAQv2 is a hybrid downscaling method that combines results from two methods – Bias Corrected Constructed Analogs – BCCA¹ and Quantile Delta Mapping – QDM²

The BCCA method is used to scale GCM data to observed data (daily gridded historical observations) from 1950 to 2010. This was done for daily maximum temperature, daily minimum temperature, and precipitation separately. This step uses both climate analogues (CA) and bias correction (BC). The climate analogue step finds observed spatial climate fields which are similar to those in the climate model. This step results in the coarse scale daily GCM data being both downscaled to a realistic fine scale and bias-corrected. However,  this can result in the removal of some of the climate change signals in the GCM data. Therefore the second step, quantile delta mapping, is used to preserve these signals in the GCM data.

As a result, the BCCAQv2 method matches the overall distribution of observations well, rather than just bias correcting to fit the mean values, which is often a more common approach. Climate model data downscaled using this method has been shown to describe the historical climate well.³

How can data downscaled using this method be used?

GCM data which have been downscaled using BCCAQv2 are useful for analysis which is specific to one variable (e.g., hottest day, maximum 1-day precipitation), and which is conducted over large spatial regions, such as the entire country. Examples of this include the use of metrics which are threshold (e.g., number of days with maximum temperature > 25°C), average (e.g., mean temperature), accumulation-based (e.g., total precipitation), or extreme events which are not tied to specific days. 

As a general rule of thumb, however, more care is necessary when assessing changes to climate metrics that relate to specific event types, locations, multi-day weather events, weather events which involve more than one variable, and in locations where there is limited input data in the target dataset (e.g., in northern locations where the observed station network is sparse).