Data assimilation

Data assimilation is a procedure that combines direct measurements, such as those taken from radiosonde balloons, with information from predictive models to give the best possible estimate of the Earth's atmosphere (its temperature or humidity profile, winds, clouds, etc) and the surface at a given time. A time-series of these snapshots of the Earth are used to study global events such as El Niño. Single snapshots are also used to initialise weather forecasting (NWP) models. Satellite data are a critical component of data assimilation systems, because they are able to observe areas that are not covered well by direct measurements, such as oceans. Much more satellite-derived information than conventional (surface and upper air) data is used to initialise current NWP models. For instance, a sequence of images from a geostationary satellite can be used to identify the movement of clouds, from which the ambient winds can be derived. These winds are said to be retrieved or assimilated from satellite measurements.

Satellite measurements are often referred to as remotely-sensed data, not just because satellites are remotely located from the weather they try to measure, but also because they do not directly measure atmospheric quantities such as temperature, humidity, and winds. Instead, satellites measure the radiation (heat) emitted by Earth's surface and its atmosphere and/or the radiation from the sun (or a power source on the satellite, as in the case of the precipitation radar on the TRMM satellite) reflected and scattered by clouds and the Earth's surface and atmosphere. Physical models must then be used to relate the satellite radiation measurements to the geophysical parameters used in weather prediction (temperature, humidity, and winds). Therefore, satellite measurements are fundamentally different from direct measurements, such as those made with instruments on balloons, that actually sample the atmosphere. The retrieval step is further complicated, because for many types of satellite measurements the instrument is not able to fully measure the vertical structure of the atmosphere. In this case, a previous estimate of the atmospheric parameters (or initial guess) is necessary to complete the retrieval.

A number of NWP centers have recently achieved remarkable success in improving their forecast skill by changing the method in which satellite data are assimilated into the forecast model. Traditionally, temperature and humidity information have been first retrieved from the satellite data and then assimilated into the NWP system. Recently, the satellite radiation measurements have been directly ingested into numerical weather prediction systems by incorporating the physical models of the transfer of radiation through the atmosphere. Incorporating such models into assimilation systems adds complexity and increases the amount of computer power needed for the assimilation system, but it has lead to improved forecasts in the short and medium ranges.