Examination of all‐sky infrared radiance simulation of Himawari‐8 for global data assimilation and model verification

Abstract

AbstractThe systematic difference between observations and simulation from weather forecast model hampers effective data assimilation and model improvement. The purpose of this study is to identify the characteristics and cause of the systematic difference or observation‐minus‐background (O − B) bias for all‐sky infrared radiances of the Himawari‐8 satellite, and propose data assimilation preprocessings and model verification. The O − B bias in cloudy scenes showed substantial negative values because of the shortage of high‐altitude clouds generated in the forecast model. Additionally, a positive bias appeared for thin ice clouds because of the excessive absorption of radiative transfer models (RTMs). These biases were traced based on a bottom‐up approach investigating individual uncertainty of RTMs, observation calibration, and the forecast model using two RTMs, reference hyperspectral sounders and synergetic measurements of CloudSat and CALIPSO. Based on these findings, data assimilation preprocessing such as quality‐control procedures excluding samples that models poorly reproduced was developed. Although the quality controls reduced the number of biased samples, non‐negligible O − B biases remained. Possible problems and treatments for the biases were discussed, including bias correction, observation error inflation, and correction of the cloud effect parameter. The O–B statistics also suggested insufficient representation of the diurnal variation in the cloud fraction in the tropics. Modified physical processes in the forecast model to increase ice clouds were tested to help improve the model bias and develop data assimilation. This trial indicated the difficulty in improving both O − B bias and variance and the necessity of adjusting the cloud effect parameters in data assimilation.