Abstract:The clear channel scheme presented by McNally and Watts (2003) is used to detect cloud radiances in GRAPES 3Dvar system measured by high spectral resolution infrared sounders. This scheme is based on the bias of simulated radiances and actual radiances, and the determination of bias is related to forecast model. Traditionally, we use subjective determination schemes to select threshold in cloud detection, such as image contrast analysis. This paper presents two new and objective schemes based on the subjective scheme. (1) Analysis of view field bias: observe the cloud top pressure under different thresholds to evaluate the rationality by analyzing the vertical distribution of bias between simulated observation and actual observation. (2) Evaluation of cloud detection index: establish a series of cloud evaluation indices to select appropriate thresholds. By combining the three methods mentioned above, we determine the IASI radiances cloud detection. The gradient threshold keeps the same value as that from European Center for it has a weak effect in cloud detection. The bias threshold is determined by a comprehensive analysis method, thus overcoming the visual error of the pure use of image. The results show that bias threshold selects 2.0 K, gradient threshold selects 0.4 K for window channels and 0.02 K for non window channels. Finally, a month long experiment shows that the cloud detection bias threshold is effective by comparing the assimilation results.