D7 Reanalysis and Integrated Products

Topic

Climate reanalyses are retrospective reconstructions of the climate system that combine historical surface and satellite observations with numerical models using data assimilation techniques. The result is a three-dimensional, temporally continuous field of variables—such as temperature, wind, humidity, and pressure—represented on a regular grid. Thus, reanalysis provides values ​​for any point and time within its coverage, even where no direct observations existed.

The current benchmark atmospheric reanalysis spans from 1940 to the present, featuring a horizontal resolution of approximately 31 kilometers and 137 vertical levels distributed throughout the atmospheric column. Independent reanalyses with comparable characteristics have also been developed by other agencies. The availability of multiple products allows for the comparison of their reconstructions; regions or periods showing the greatest differences indicate higher uncertainty regarding the reconstructed climate state.

Ocean reanalyses apply the same principle to the ocean. They combine numerical models with available observations, including temperature and salinity profiles obtained from autonomous floats and oceanographic cruises. This integration yields a continuous three-dimensional representation of temperature, salinity, and currents, encompassing regions that lack direct measurements.

The fundamental technique enabling the creation of these products is data assimilation. At each stage, the model-generated forecast—known as the "background"—is combined with available observations to produce an improved estimate of the system's state, known as the "analysis." Modern methods aim to balance both sources of information by accounting for their respective uncertainties.

One of the primary techniques is four-dimensional variational assimilation. This method minimizes a cost function that penalizes both the difference between the analysis and the background state and the difference between the analysis and the observations. Furthermore, it requires the resulting evolution to be consistent with the model equations throughout the entire time window under consideration. In this way, an isolated and unreliable observation does not force the analysis to deviate excessively from the state provided by the model: the weight assigned to each data point depends on its reliability.

Alongside reanalyses, there are integrated precipitation products that employ a similar logic for combining information. These merge measurements obtained via satellite—using radar and passive microwave radiometers—with records from surface rain gauges. Satellites offer extensive spatial coverage, while rain gauges provide precise point measurements. Fusion algorithms leverage both advantages to produce global precipitation estimates with temporal resolutions as fine as half an hour and spatial resolutions on the order of one-tenth of a degree—characteristics that neither source could provide on its own.

Another type of integrated product consists of databases of climate extremes indices. These utilize previously homogenized daily observations to calculate standardized indicators, such as temperature percentiles, the number of frost days, heatwave duration, and peak precipitation levels. The preliminary homogenization process aims to generate time-comparable series before using them to characterize climate changes and extremes.

Taken together, atmospheric and oceanic reanalyses, integrated precipitation products, and extremes index databases transform observations that are scattered, incomplete, and derived from diverse instruments into coherent representations of the climate system. Reanalyses use models and data assimilation to reconstruct continuous fields; precipitation products merge satellite and ground-based data; and index databases process homogenized series to describe extremes. Their shared purpose is to provide a spatially and temporally complete, consistent view of the climate that no single, isolated observation could offer.

ID:906