3.4 Reanalysis data Article Swipe
YOU?
·
· 2020
· Open Access
·
· DOI: https://doi.org/10.1051/978-2-7598-1849-5.c023
· OA: W4234722406
What are reanalysis data?Compare weather-or climate-model output to observations: Not only are the model outputs regularly gridded in time and space but are also complete.The availability of observations depends on the coverage, in space and time, of global observing systems, which includes weather stations, radiosondes, airplanes, and instruments onboard satellites.Despite this wealth of data, the coverage is not complete and data quality may vary with time.Some form of data selection and interpolation, or other processing, can bring observations into a uniform shape, which is appealing from a data analysis point of view.Purely mathematical interpolation can do this, but does not provide any extra information, as it is unphysical.In reanalysis, physical models are used to provide the extra information, resulting in a physically consistent description.Models used for reanalysis are meteorological numerical weather prediction (NWP) models that assimilate available observations and ensure that the numerical solution differs as little as possible from the observations where these are present, and offers physically consistent values for data sparse regions.The models start at a state given by previous solutions, calculate one time step forward, compare the solution to observations that are relevant for the time, calculate the error, go back and modify the starting condition, and so on: by iteration, a solution (the 're-analysis') is achieved in which the model and the observations are as consistent as possible to within their respective errors.The process is then repeated for the next time step.Reanalysis grew out of the NWP field since NWP models base their forecasts on current observations and a physical model.As part of the data assimilation, a physics-based quality control is performed.Reanalysis models use similar model software but assimilate historical data -weather station
