This dataset contains monthly-mean radiance observations from the NASA AIRS instrument (Chahine, 2006) on a regular, lat-lon grid. Data are available from 22 years of observations from September 2002 through August 2024. Profiles with non-nominal quality flags are excluded. Data are separated by zenith viewing angle, and orbit type (ascending vs. descending), and the number of aggregated observations is included to allow for tests of statistical significance, assessment of data outages, etc. To distinguish between clear and cloudy scenes, we identify clear-sky observations following the methods of DeSouza-Machado et al., 2025. Namely, profiles with the top 10% of radiances in the 1231 cm-1 window channel are labeled as clear-sky. DeSouza-Machado et al., 2025 show that while this methodology is sensitive to cloud contamination in regions with high cloud cover and frequent temperature inversions, it provides a good estimate of clear-sky conditions with global coverage. Clear-sky profiles are identified separately for each 16-day AIRS orbit cycle and are then aggregated to monthly mean fields. Spatial grid spacing is 5 degrees in longitude, with variable latitude spacing to give a roughly equal number of observations in each gridcell. 470 spectral channels on the AIRS L1C spectral grid are included here. Those channels were identified as the most suitable for climate analysis studies in Strow et al., 2020 (see Section 4.2). A list of channels can be found at https://doi.org/10.5281/zenodo.3878740. Each "chunk" file of MAR contains 10 spectral channels from this 470 channel subset (e.g. chunk0 contains the first 10 channels, chunk1 the next 10, etc). Once unzipped, these can be easily concatenated and operated on. The included python script (MARexample_AIRSobs_averaging.py) provides example code for loading and processing MAR. More extensive examples can be found in code documentation for Shaw et al., 2026 (https://doi.org/10.5281/zenodo.18167283). Data are available in the Zarr data format to be efficiently accessed and reconfigured for different user purposes. Each spectral channel is ~370mb and the entire dataset is ~180GB. Zarr format 2 is used to remain compatible with both Zarr 2 and 3 versions. These data were produced for Current and Future Changes in Earth's Outgoing Infrared Spectrum (Shaw et al., 2026), which serves to document them. Per the license, please appropriately acknowledge Shaw et al. (2026) and DeSouza-Machado et al. (2025) if you use these data. References: Chahine, M. T., and Coauthors (2006). AIRS: Improving Weather Forecasting and Providing New Data on Greenhouse Gases. Bull. Amer. Meteor. Soc., 87, 911–926, https://doi.org/10.1175/BAMS-87-7-911 DeSouza-Machado, S., Strow, L. L., Tangborn, A., Huang, X., Chen, X., Liu, X., Wu, W., and Yang, Q. (2018). Single-footprint retrievals for AIRS using a fast TwoSlab cloud-representation model and the SARTA all-sky infrared radiative transfer algorithm. Atmos. Meas. Tech., 11, 529–550, https://doi.org/10.5194/amt-11-529-2018 DeSouza-Machado, S., Strow, L. L., & Kramer, R. J. (2025). Geophysical trends inferred from 20 years of AIRS infrared global observations. Journal of Geophysical Research: Atmospheres, 130, e2025JD043501. https://doi.org/10.1029/2025JD043501 Strow, L. L. and DeSouza-Machado, S. (2020). Establishment of AIRS climate-level radiometric stability using radiance anomaly retrievals of minor gases and sea surface temperature. Atmos. Meas. Tech., 13, 4619–4644, https://doi.org/10.5194/amt-13-4619-2020 Shaw, J.K., Kay, J.E., DeSouza-Machado, S., Turner, D.D., Strow, L. L. (2026). Current and Future Changes in Earth's Outgoing Infrared Spectrum. Geophysical Research Letters, https://doi.org/10.1029/2026GL121893 (in press)