These surface spectral radiative kernels are computed by Xianglei Huang's group at the University of Michigan, using one year of 6-hourly profiles from MERRA-2 and MODTRAN 5.2.0.0 radiative transfer model. This product is an outcome of an NASA MEASURES project (80NSSC24M0014) awarded to the University of Maryland, Baltimore County, with a subcontract to the University of Michigan. The surface radiative kernels are defined as the sensitivity of downward longwave flux at the surface to the atmospheric states (T, Ts, q and CO2). The datasets contain 96 NetCDF files: 12months*4variables(T,q,Ts,CO2)*2types(all-sky and clear-sky)Total size: 38G T and q kernels have 4 dimensions: lat, lon, level, frequency Ts and CO2 have 3 dimensions: lat, lon, frequency Resolution:181 lat grids: -90 to 90, every 1 degree144 lon grids: -180 to 177.5, every 2.5 degree17 pressure layers, lower boundary: 1050, 950, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 70, 50, 30, 20 hPa upper boundary: 950, 925, 850, 700, 600, 500, 400, 300, 250, 200, 150, 100, 70, 50, 30, 20, 10 hPa440 frequency grids: 0-2195 cm-1, every 5 cm-1. A sample code, example.m is an example for using the dataset. If you have any questions about the dataset, please contact Xiuhong Chen at xiuchen@umich.edu or Xianglei Huang at xianglei@umich.edu. If you use these datasets for publication, please cite Chen, X. H and X. L. Huang, 2025: MERRA-2 based surface spectral radiative kernels. https://doi.org/10.5281/zenodo.17675159