Key features and evaluation of version 3 of Environment and Climate Change Canada’s Regional Surface and Precipitation Reanalysis

Using operational surface and precipitation data assimilation and forecast systems along with surface observations, Environment and Climate Change Canada (ECCC) developed a regional surface and precipitation reanalysis covering North America at a 10km spatial resolution and spanning nearly four decades. An offline precipitation reanalysis is also provided for the same reanalysis system, allowing assimilation of additional precipitation observations. While three reanalysis versions have been already released and can be found on the Canadian Surface Prediction Archive (CaSPAr; https://caspar-data.ca) platform, the current presentation focuses on the soon to be released, latest version 3.

The reanalysis follows a sequential modeling hierarchy. First, an existing global atmospheric reanalysis initializes a global numerical weather prediction (NWP) called the Global Deterministic Reforecast System (GDRS). Outputs from GDRS are, in turn, dynamically downscaled to a 10-km grid resolution by the Regional Deterministic Reforecast System (RDRS), which is coupled to the Regional Surface Assimilation System (RSAS) in an assimilation cycle leading to the production of the surface and precipitation reanalysis. All components of the system rely on ECCC operational configurations: GDRS and RDRS are based on the Global Environment Multiscale (GEM) model from the Global and Regional Deterministic Prediction Systems (GDPS and RDPS), while RSAS is built from the Canadian Land Data Assimilation System (CaLDAS) and the Canadian Precipitation Analysis system (CaPA) configurations used to initialize the High Resolution Deterministic Prediction System (HRDPS). Version 3 of the North American Surface and Precipitation reanalysis introduces significant updates from its predecessors, version 2.1. Noteworthy, changes include, among others, shifting to ERA-5 from ERA-Interim for initialization, upgrading GEM version from 4.8 to 5.2 (which represents more than 5 years of model development by the Meteorological Research Division (MRD) of ECCC), improving the assimilated surface observation datasets, improving the snow representation in the land data assimilation system, correcting a cold bias in the Canadian Western Coastal Mountain ranges, and extending temporal coverage to 1980-2023.

This presentation introduces key features and methodology of the third version, along a summary of its main differences from version 2.1. Then, an evaluation of the main variables (i.e., surface temperature and dew point, precipitation and snow) of the third version of the reanalysis is presented, comparing it against independent observation datasets, previous version of the reanalysis and other mainstream reanalysis such as ERA-5 Land and MERRA-2. Finally, successful applications of the dataset to various problems are briefly presented, including using the reanalysis as input of more advanced surface and hydrology models and products, extreme event characterization for flood plain modeling, their use as a benchmark for future climate projections, and their use to train artificial intelligence approaches for hydrological prediction.