摘要
Three kinds of the widely-used cloudiness parameterizations are compared with data produced from the cloud-resolving model(CRM) simulations of the tropical cloud system. The investigated schemes include those based on relative humidity(RH), the semi-empirical scheme using cloud condensate as a predictor, and the statistical scheme based on probability distribution functions(PDFs). Results show that all three schemes are successful in reproducing the timing of cloud generation, except for the RH-based scheme, in which low-level clouds are artificially simulated during cloudless days. In contrast, the low-level clouds are well simulated in the semi-empirical and PDF-based statistical schemes, both of which are close to the CRM explicit simulations. In addition to the Gaussian PDF, two alternative PDFs are also explored to investigate the impact of different PDFs on cloud parameterizations. All the PDF-based parameterizations are found to be inaccurate for high cloud simulations, in either the magnitude or the structure. The primary reason is that the investigated PDFs are symmetrically assumed, yet the skewness factors in deep convective cloud regimes are highly significant, indicating the symmetrical assumption is not well satisfied in those regimes. Results imply the need to seek a skewed PDF in statistical schemes so that it can yield better performance in high cloud simulations.
Three kinds of the widely-used cloudiness parameterizations are compared with data produced from the cloud-resolving model （CRM） simulations of the tropical cloud system. The investigated schemes include those based on relative humidity （RH）, the semi-empirical scheme using cloud condensate as a predictor, and the statistical scheme based on probability distribution functions （PDFs）. Results show that all three schemes are successful in reproducing the timing of cloud generation, except for the RH-based scheme, in which low-level clouds are artificially simulated during cloudless days. In contrast, the low-level clouds are well simulated in the semi-empirical and PDF-based statistical schemes, both of which are close to the CRM explicit simu- lations. In addition to the Gaussian PDF, two alternative PDFs are also explored to investigate the impact of different PDFs on cloud parameterizations. All the PDF-based parameterizations are found to be inaccurate for high cloud simulations, in either the magnitude or the structure. The primary reason is that the investigated PDFs are symmetrically assumed, yet the skewness factors in deep convective cloud regimes are highly significant, indicating the symmetrical assumption is not well satisfied in those regimes. Results imply the need to seek a skewed PDF in statistical schemes so that it can yield better performance in high cloud simulations.
基金
supported by the National Basic Research Program of China(Grant Nos.2014CB441202,2013CB955803)
the National Natural Science Foundation of China(Grant Nos.41305102,91337110)
the Strategic Priority Research Program of the Chinese Academy of Sciences(Grant No.XDA11010402)
the Joint Center for Global Change Studies(Grant No.105019)
