Propagation of uncertainty in Earth Observation

This figure represents the way in which uncertainty flows and originates in environmental and climate data derived from Earth Observation (satellite data). The nomenclature of "satellite processing levels" is used. At each stage of transformation, uncertainty propagates (flows) from lower level data to the higher (more processed) product. Each stage of transformation (such as retrieval of a geophysical quantity from radiances, or aggregation of data to gridded form or summary indices) involves assumptions, approximations and/or parameters that also inject new sources of uncertainty. Some transformations, such as averaging, also reduce components of uncertainty by the cancellation of some errors. This means that sources of uncertainty that are minor (in terms of magnitude) at lower levels can become dominant in higher level products (as discussed here: http://www.fiduceo.eu/content/why-worry-about-all-sources-errors). Ideally in science, all measured values should have an associated estimate of uncertainty, but the flow of uncertainty shown in the figure is not systematically understood and presented at the current time.