【摘要】 无论是全球气候变化研究还是降水定量预报研究,都涉及如何将区域尺度气象因子向局部尺度聚解(Downscaling)的问题。提出了两种各有特色的聚解,一种是与天气动力学模式相联系的灰色系统模型方法。在宏观或中观尺度的大气动力学模型和地面局部尺度降水变化之间,嵌入一个反映区域尺度气象因子向局部尺度聚解的灰色转化器。运用我们研究的灰变量运算法则,能够将其简化为一个确定性的聚解方程和反映聚解转化不确定性的灰区间半径系统方程。通过系统识别可分析聚解过程中的确定性规律和不确定性风险问题。另一种是模式识别的方法。它的特点是在利用模式识别理论,分别辩识天气动力学模式的特征向量和地面降水模式的特征向量。然后,设法建立二者多变量的联系。给出了两种方法在湖北省降水定量预报的一个应用实例。对动力学方法、黑箱子的模式识别方法和介于两者之间的与动力学模式耦合的灰色系统方法进行了比较与评述,指出目前需进一步研究的问题。更多还原

【Abstract】 This paper introduces two daily rainfall downscaling methods from regional scale to local scale. One is called as the Grey System Approach. This model is formulated by a grey differential model, coupled with the Finite Domain Mesoscale Atmospheric Model (FDMAM) as a means for predicting daily precipitation in high resolution. A new operation law of grey number was developed that overcomes some of disadvantages of the present interval operations. In the light of operation principles, an original grey differential equation could be reduced as two equivalently sub-models. One describes linkage of averaged state of downscaling process. The other expresses uncertainty in downscaling. The grey system approach improves upon previous prediction model by allowing uncertainty information to be directly communicatied into the grey series modeling and prediction. The second method is named as a pattern Recognition approach. Using the KarhunenLoeve (K-L) orthogonal functional series, rainfall features could be indentified. Linking the feature vectors of atmospheric patterns with that of the time-space distribution of daily precipitation by a simple multidimensional regressive approach. These two methodologies have verified by daily precipitation observed at rain gauging stations during the period of heavy rain and the Heavy Rainfall events of 1996 in Hubei province, China. Comparisons of three downscaling methods were made in terms of several assessment criteria.更多还原