Earth observation data assimilation is the use of observations together with short-range model forecasts to determine a best possible estimate of the state of a system and its uncertainties.1

ECMWF describes the procedure as typically sequential: a previous model forecast is compared with newly received observations, the model state is updated, and a new forecast is started. The update step is called the analysis and the short forecast used to produce it is called the background.1 That description is given for atmospheric forecasting, and the sources reviewed do not describe ocean, land or other assimilation systems.1

NASA lists model output and results from analyses of lower-level data as its highest processing level.2 The NASA page does not name assimilation for that level, so the link between assimilation products and that level is an inference and not a statement of the source.2

ESA describes a Digital Twin Earth as a digital replica of the planet that is constantly fed with Earth observation data, combined with in situ measurements and artificial intelligence.3 The article reports ideas and precursor activities and does not establish that assimilation is the mechanism of any particular twin, so the connection should be read as a shared use of observations and models.3 Assimilation is a method of combining data with a model and is not the same as the observations it uses or the forecast it starts.

Footnotes

  1. ECMWF, “Data assimilation”. https://www.ecmwf.int/en/research/data-assimilation ↩ ↩2 ↩3

  2. NASA Earthdata, “Data Processing Levels”. https://www.earthdata.nasa.gov/learn/earth-observation-data-basics/data-processing-levels ↩ ↩2

  3. ESA, “Working towards a Digital Twin of Earth”, 14 October 2021. https://www.esa.int/Applications/Observing_the_Earth/Working_towards_a_Digital_Twin_of_Earth ↩ ↩2