Hyperspectral infrared satellite observations have been collected in large numbers since the year 2002. The current hyperspectral sounders, AIRS on Eos-Aqua, IASI on MetOp-A and -B, and CrIS on Suomi NPP, were preceded by instruments flown as early as the 1960s. The data collected then, as well as today, have so far been largely under-utilized in atmospheric reanalyses, or at worst ignored completely. Yet, these data contain spectrally detailed information about our atmosphere's vertical structure and its constituents, and great potential to serve as stable references.
The first part of this talk will review how reanalyses are conducted, thereby explaining the current under-utilization mentioned above. Some reanalyses involve (century-)long integrations such as the NOAA 20th century reanalysis or the ECMWF 20th century reanalysis (ERA-20C). Other reanalyses only cover the recent satellite era, such as ERA-Interim. Special attention will be given to the topic of observation biases and the evolution of modern reanalyses into near-real-time climate monitoring systems. These latter reanalyses create emerging new needs on the data side.
In a second part of the talk we will take a look at data from the InfraRed Interferometer Spectrometer (IRIS). This was an early version of the instruments later carried onboard Voyager-1 and -2 spacecrafts to explore the planetary atmospheres of our solar system. The IRIS experiments employed robust instrument design from the 1960s. They were flown on the Nimbus-3 and Nimbus-4 satellites. We will focus on the Nimbus-4 IRIS data, between April 1970 and January 1971. Radiance data from this early hyperspectral 862-channel Michelson interferometer have recently been rescued from magnetic tapes by the NASA Goddard Earth Sciences Data and Information Services Center (GES DISC). Using modern radiative transfer tools, namely EUMETSAT NWP-SAF RTTOV, we will show comparisons of these radiance data with ECMWF reanalyses.
Although past behavior may only serve as a guide to future expectations, the IRIS example suggests that the current inconvenience of storing all hyperspectral data, although significant at collection time (when data volumes are challenging infrastructures, bandwidth and storage), may gradually decrease over time (as data storage and transfer solutions become more powerful while data volumes remain constant); in parallel, one may expect growing benefits to be gained from these data, thanks to an increasingly long data record and more powerful computers and better models to interpret the information, and their potential role as calibration references.
Overall this example shall demonstrate the importance of retaining as much as possible all of the environmental information collected by hyperspectral sounders data and meta-data. These could prove as invaluable information at a later time to determine subtle structural changes that are taking place over long time scales in our atmosphere.