| description abstract | Spring rainfall is important for agriculture and economics in North China (NC). Thus, there is an imperative need for accurate seasonal prediction of the spring precipitation. This study implements a novel rainfall framework to improve understanding of NC spring rainfall. The framework is built based on a three-cluster normal mixture model. Distribution parameters are sampled using Bayesian inference and a Markov chain Monte Carlo algorithm. The probability behaviors of light, moderate, and heavy rainfall events can be reflected by the three rainfall clusters, respectively. Analysis of 61-yr data indicates that moderate rainfall makes the largest contribution (67%) to the total rainfall amount. The moderate rainfall intensity is mainly influenced by the sea surface temperature anomaly (SSTA) in the previous season over the equatorial eastern Pacific, and rainfall frequency is influenced by geopotential height anomaly in the mid- to high latitudes in spring. It is also found that more extreme precipitation events can be observed in the spring following an eastern Pacific El Niño event in the previous autumn and winter. Based on these relationships, we develop a multiple linear regression model. Hindcasts for spring precipitation using the model indicates that its anomaly correlation is 0.48, significant at the 99% confidence level. The result suggests that the newly developed model can well predict spring rainfall amount in NC. | |
