Sensitivity of Tropical Tropopause Layer Cirrus Prediction in GRAPES Global Forecast System

Abstract

Abstract A warm bias with a maximum value of over 4 K in the tropical tropopause layer (TTL) is detected in day-5 operational forecasts of the Global/Regional Assimilation and Prediction System (GRAPES) for global medium-range numerical weather prediction (GRAPES_GFS). In this study, the predicted temperature changes caused by different processes are examined, and the predicted cloud fractions are compared with the European Centre for Medium-Range Weather Forecasts ERA5 reanalysis data. It is found that the overprediction of the TTL cirrus fraction contributes to the warm bias due to cloud-radiative heating. The interactions among the ice nucleation, deposition/sublimation, and the large-scale condensation together determine the results of the TTL ice crystal content prediction. Moreover, a range of sensitivity experiments show that the TTL ice crystal content prediction is sensitive to the threshold relative humidity over ice (RHi) in the ice nucleation process. Then the uncertainties of the formulas for saturation vapor pressure over ice at very low temperatures are discussed. The RHi calculated based on the Magnus–Tetens formula is up to 10% higher than that based on the Goff–Gratch formula. As the Goff–Gratch formula is applicable over a broader range of 184–273 K, it is more suitable for the cold TTL. When the Goff–Gratch formula rather than the Magnus–Tetens formula is used in the microphysics scheme, the TTL cirrus forecasts are improved greatly, and the warm bias disappears completely. After investigating the interplay of the dynamical, microphysical, and radiative processes, we find a positive feedback mechanism that exacerbates the TTL cirrus prediction error.