Analysis of MTG-IRS observations and general channel selection for numerical weather prediction models

The Infrared Sounder (IRS) is an infrared Fourier transform spectrometer that will be on board the Meteosat Third Generation series of the future European Organization for the Exploitation ofMeteorological Satellites geostationary satellites and will have a unique four- dimensional look at the atmosphere. After its planned launch in 2024, it will be able to measure the radiance emitted by the Earth at the top of the atmosphere using 1,960 channels in two spectral bands between 680-1,210 cm(-1) (long-wave infrared) and 1,600-2,250 cm(-1) (mid-wave infrared) at a spectral sampling of 0.6 cm(-1). It will perform measurements over the full Earth disk with high spatial and temporal resolution of 4 km at nadir and 30 min over Europe. Thus, the huge amount of data from IRS will present challenges, particularly in data transmission, data storage, and assimilation into numerical weather prediction (NWP) models. To reduce the volume of data, various methods are available, including spatial sampling, principal component analysis, and channel selection. The latter technique will be discussed in this paper by proposing general channel selection to provide NWP models. The objective of this selection is to improve essential variables for NWP such as temperature, humidity, skin temperature, and ozone. This work has required the development of a large observation database and takes into account the main developments in assimilation techniques, including the use of full observation-error covariance matrices or the assimilation of ozone in global models, for example. This study performs a specific analysis of the sensitivity of IRS observations and proposes a first general selection of 300 channels for NWP models. This selection allows us to reduce the analysis error in the troposphere by 48% in temperature, 65% in humidity, and 17% in ozone.