Reanalysis vs Forecasting

Understanding ERA5 requires distinguishing reanalysis from forecasting. Both use numerical models, but their purpose and methodology differ fundamentally.

1. Forecasting

Definition: A forecast is a forward-in-time prediction of the atmospheric state, starting from the best estimate of current conditions.

Mathematical representation:

$$ \text{State}(t_0) \xrightarrow{\text{numerical model}} \text{State}(t_0 + \Delta t) $$

Key characteristics:

Uses real-time initial observations
Model is run forward in time
Errors grow with lead time due to chaos
Once issued, a forecast does not change

Example: A 5-day weather forecast initialized at 00 UTC today.

2. Reanalysis

Definition: A reanalysis reconstructs past atmospheric conditions using a fixed modern model and assimilating all available historical observations.

Conceptual representation:

$$ \text{Observations (past + future)} \Rightarrow \text{Best estimate of past state} $$

Key characteristics:

Retrospective: covers decades
Uses a fixed model version for temporal consistency
Assimilates a wide range of historical data
Produces gap-free, physically consistent datasets

Example: ERA5 reconstructs global atmospheric conditions from 1940–present.

3. Comparison Table

Aspect	Forecasting	Reanalysis (ERA5)
Time direction	Forward	Retrospective
Goal	Predict future atmospheric states	Reconstruct past atmospheric states
Observations	Real-time, incomplete	All available historical observations
Model version	Changes over time	Fixed for entire dataset
Consistency	Varies between years	High temporal consistency
Error growth	Increases with lead time	Minimized by assimilation
Updates after release	No	No (dataset frozen)
Best use	Weather prediction	Climate studies, station comparisons, physical process analysis

4. Why ERA5 is Reanalysis

ERA5 is not a forecast because it does not predict future states. Instead:

It uses a single, state-of-the-art ECMWF model
Assimilates the following data sources:

Surface stations
Radiosondes
Satellites
Aircraft
Buoys
Reprocessed historical datasets

Provides a best-guess atmospheric history

For example, ERA5 temperature for 2015 is a reconstruction, not a forecast issued in 2015.

5. One-Sentence Summary

Unlike numerical weather forecasts, which predict future atmospheric states from real-time initial conditions, ERA5 is a reanalysis product that reconstructs past atmospheric conditions using a fixed modern model and comprehensive historical observations, providing temporally consistent datasets suitable for climate analysis.