Optimized Wavelength Sampling for Thermal Radiative Transfer in Numerical Weather Prediction Models

In the thermal spectral range, there are millions of individual absorption lines water vapor, CO2, and other trace gases. Radiative transfer calculations wavelength-integrated quantities, such as irradiance heating rate, computationally expensive, requiring a high resolution for accurate numerical weather prediction climate modeling. This paper introduces method that could highly reduce cost integration in spectrum by employing an optimized wavelength sampling method. Absorption optical thicknesses various gases were calculated from HITRAN 2012 spectroscopic dataset using ARTS line-by-line model input to fast Schwarzschild radiative model. Using simulated annealing algorithm, different sets wavelengths corresponding weights identified, which allowed integrated quantities be computed weighted sum, reducing computational time several orders magnitude. For each set wavelengths, lookup table, including cross-sections, is created can applied any atmospheric setups it was trained. We table calculate irradiances rates large profiles ECMWF 91-level short-range forecast. Ten nodes sufficient obtain within average root mean square error (RMSE) upward downward radiation at height below 1 Wm?2 while 100 RSME 0.05 Wm?2. The applicability this confirmed clear conditions exemplary cloud 3.2 km height. Representative gridpoints (REPINT) available parameterization libRadtran package, where used efficient molecular variety solvers.