基于分层贝叶斯框架的青藏高原降水相态识别方法研究

The detection and retrieval of snowfall in high-cold regions have always been one of the most significant challenges and topical issues of the precipitation studies over the world, in which determining the precipitation phase based on environmental variables is the core constraint in its long-term development. In this proposal, on basis of the Hierarchical Bayesian framework (HBF), we develop a Markov Chain Monte Carlo (MCMC) sampling derived scheme of precipitation phase that incorporates the multi-environmental parameters across the Tibetan Plateau (TP). This study includes three sections: (1) identifying and testifying the key surficial and atmospheric parameters which potentially attribute to the discrimination of solid-liquid precipitation in TP; (2) developing a parameterized scheme of precipitation phase based on HBF and multi-environmental variables, and thus quantifying the conditional probability of solid precipitation and its error distribution; (3) applying the proposed precipitation phase scheme for multi-source (e.g. gauge and satellite) precipitation observations over the TP. In summary, this proposal will provide theoretical and methodological references for high-accurate discrimination of precipitation phase in high-cold regions, and be helpful for updating the precipitation database in the Third Pole, as well as contributing to high-cold region’s snowfall retrieval algorithm for Global Precipitation Measurement (GPM) mission.

高寒区降雪的探测与反演一直是国际上降水研究的前沿和难点，其中如何基于环境变量准确地估测目标区降水相态是其关键性制约因素。本项目拟选择具有代表性的青藏高原作为研究区域，引入分层贝叶斯框架，构建基于马尔可夫链蒙特卡洛混合采样策略的多环境因子驱动的降水相态识别方法，将开展三个方面的研究：1) 诊断并检验影响青藏高原降水相态特征的关键地表、大气等环境因子；2) 基于分层贝叶斯框架建立多环境因子驱动的青藏高原降水相态识别方法，定量分析目标区降雪发生概率及误差特征；3) 本项目设计的降水相态识别方法在青藏高原地-空多源降水观测中的应用。该研究成果将为高寒区降水相态的高精度识别提供理论和方法参考，有助于进一步完善第三极地区降水资料库，并对新一代全球降水观测计划在高寒区的遥感降雪反演算法做出中国特有的贡献。

高寒区降雪的探测与反演一直是国际上降水研究的前沿和难点，其中如何基于环境变量准确地估测目标区降水相态是其关键性制约因素。本项目选择具有代表性的青藏高原作为研究区域，引入分层贝叶斯框架，构建基于马尔可夫链蒙特卡洛混合采样策略的多环境因子驱动的降水相态识别方法。. 本项目取得的重要结果包含以下几个方面：. 1. 通过收集潜在影响青藏高原降水相态的地表、大气等环境变量，建立了1960-1979年近20年青藏高原降水类型及相关辅助因子的资料库；. 2. 通过相关分析阐明了青藏高原影响降水相态特征的关键环境变量，认为湿球温度、相对湿度、气候区划是最为关键的三大指标，并且湿球温度的影响最直接，而且在不同气候区的两大关键环境变量的影响程度有所差异；. 3. 基于分层贝叶斯框架构建了高寒区多环境变量驱动的降水相态识别算法，进一步通过青藏高原10%独立地面观测资料检验该算法的适用性，初步验证其模型准确率超过85%。. 该研究成果将为高寒区降水相态的高精度识别提供理论和方法参考，有助于进一步完善青藏高原降水资料库，并对新一代全球降水观测计划在高寒区的遥感降雪反演算法做出中国特有的贡献。