Earth Observation data are usually made available in different processing levels. The processing level is a mean of describing how much the raw data have been processed toward an informational geophysical product. The degrees of data processing usually follow a numerical hierarchy and typically range from Level 0 (less processed) up to Level 4 (highly processed). They characterize the type of data processing that has been performed between the raw data and the current product.
A first effort for providing standard definitions of different processing levels has been made in the 1980s by the Committee on Data Management and Computation (CODMAC) of the National Research Council (NRC). CODMAC identified eight levels of processing, applicable for all space science data. Starting with the raw data at level 1, the degree of processing and complexity of the data increased at each new level. Level 2 describes edited data, corrected for obvious instrumentation errors and tagged with acquisition time and location; Level 3 stays for calibrated data where values are proportional to a specific physical unit. Level 4 represents resampled data, Level 5 derived data, where specific geophysical information has been retrieved and mapped based on the original data. Level 6 represents all ancillary data (i.e. instrument data) that are necessary for the previous steps of calibration and resampling. Level 7 describes so called correlative data: not directly belonging to the original data, those data represent all other science data that where necessary for the interpretation of the original spaceborne dataset. Finally, Level 8 are user description, i.e. documentation of the data.
Concerning spaceborne image data, both optical and radar, an adaptation of these original levels has been made from NASA and NOAA that is used for the main current spaceborne missions, including the Copernicus program. Whereas specific adaptations may arise for specific sensors and sensor types, there are five principal processing levels. Level 0 represents the raw data that have just been edited for the correction of artifacts. Level 1 data are Level 0 data with additional annotations regarding time and geolocation information, radiometric and geometric calibration coefficients (for example Top of Atmosphere data for optical imagery). Level 2 data are already radiometrically and geometrically calibrated and represent physical variables (for example Bottom of Atmosphere data for optical imagery). Level 3 data correspond to derived variables and information (e.g. land cover) with completeness and consistency information, e.g. quality flags. Level 4 represent higher level data resulting from modelling or more complex analysis of the data with additional ancillary information.
For many applications and users, so called analysis ready data (ARD data) are required. These usually correspond to Level 2 data that have already been pre-processed in order to retrieve the physical information and can be further analyzed for the specific thematic application.