Reanalysis Data

Reanalyses are gridded climate datasets built by running a fixed (“frozen”) weather model with an unchanging data-assimilation method across past decades. Every x hours, the system ingests all observations available at that time to reconstruct the atmosphere and surface. Because the model and assimilation stay constant, the result is a dynamically consistent estimate at each time step. What does change is the observing system feeding the model (e.g., radiosondes, satellites, buoys, aircraft and ship reports) which inevitably evolves over time.

The “Copernicus ECMWF” YouTube channel provides a short video introduction to reanalysis data:

More information can be found on the ECMWF web site What is climate reanalysis?.

Dataset Descriptions

This Document provides guidance on:

  • The Canadian Surface Reanalysis (CaSR)
  • The land component of the fifth generation of European ReAnalysis (ERA5-Land)

Strengths and Limitations

Key Strengths of Reanalysis Data

Strength Description
Spatial coverage Broad geographic coverage, including sparse/remote areas.
Continuity Multivariate, spatially and temporally complete: Every grid cell has a full set of variables at every time step.
Physical and dynamic coherence. Physically and dynamically coherent fields as variables evolve together according to atmospheric physics, so derived metrics are internally consistent.
Straightforward processing Data are shared in standard, well-documented formats on regular grids.

Key Limitations of Reanalysis Data

Limitation Description
Potential discontinuities Changing observing systems and biases can create spurious variability/trends
Spatial and temporal inconsistencies Reliability of the reanalysis varies by location, era, and variable.
Uncertainties largely unknown. Uncertainty is difficult to evaluate.

Expert Guidance

Reanalysis data are best viewed as physically consistent reconstructions, not the literal “true state of the atmosphere.” They blend incomplete, error-prone observations with imperfect models through complex data-assimilation methods. As a result, their limitations largely reflect information gaps: the atmosphere has never been fully observed; observations carry errors that are not always well characterized; models and assimilation schemes have shortcomings; technical mistakes can occur; and computational limits constrain spatial and temporal resolution. And caution is needed for low-frequency variability and trends: shifts in the observing system over time can imprint artificial steps or drifts in reanalysis series.

To assess uncertainty for a given variable, a user should consider how directly that variable is observed, how observational coverage varies across space and time (including historical changes), and how well the model represents the processes controlling that variable. These factors will determine the confidence one ought to have in this variable and any derived climate indicator. However, most users will not be able to complete this assessment due to limited expertise, time or technical constraints, or lack of access to needed information.

As such, it is recommended that users validate the reanalysis to an independent source over a recent multi-year period (e.g. nearest station). Users can also get an estimate of the uncertainty by validating their data using a second reanalysis product. As a rule of thumb, regions with fewer observations will tend to have larger errors. This includes Canada’s North, mountainous regions (BC), coasts and offshore (Great Lakes/Atlantic). That being said, in many of these cases, reanalyses will still provide the best available estimate of the recent climate.

References

Mankin, J., Lesk, C., Atmospheric Research Staff (Eds)., N.C. for, 2023. The climate data guide: Making sense of data from land surface models (LSMs) [WWW Document]. UCAR Climate Data Guide. URL https://climatedataguide.ucar.edu/climate-data/making-sense-data-land-surface-models-lsms (accessed 8.14.2025).
Schneider, D.P., Deser, C., Fasullo, J., Trenberth, K.E., 2013. Climate data guide spurs discovery and understanding. Eos, Transactions American Geophysical Union 94, 121–122. https://doi.org/10.1002/2013EO130001