融合时序特征的湖北省降水空间分布估算

以湖北省2006～2010年的降水量为基础数据源，通过对降水数据的分析，将省内75个站点的时间维降水序列变差函数值作为降水强度的影响因素，引入到降水量空间预测的多元地理统计中，使用时序协同克里格方法构建降水时序特征与降水量之间的数学关系模型，对湖北省月平均降水数据进行空间插值模拟。通过分析不同时间尺度下的交叉检验统计量，选择最优的时间维变差函数计算尺度。结果显示，融合时序特征的协同克里格法对降水量的空间插值较普通克里格法具有更好的插值效果，因而更适合历史资料丰富地区的降水空间分布估算

SPATIAL DISTRIBUTION OF PRECIPITATION ESTIMATION BASED ON CHARACTERISTICS OF TEMPORAL SEQUENCE IN HUBEI PROVINCE

The spatial distribution of precipitation estimation is the basis of the meteorological service. Inadequate of the rain gauge observation site do influence the precision of the precipitation distribution estimation. In order to minimize the impact, it is necessary to introduce the precipitation factors for auxiliary operation. But due to reasons such as confidentiality, a lot of impact factors are difficult to obtain for ordinary researchers. In this paper, the problem was solved by using the Temporal Sequence co Kriging method, which could mine temporal character from the historical precipitation data. This paper chose the precipitation data from 2006 to 2010 in Hubei Province as the research object. Through the analysis of the rainfall data, the value of Variation Function for every rain gauge as the factor that influenced rainfall intensity was introduced to the precipitation distribution estimation of multiple geographic space in statistics, and the Temporal Sequence co Kriging method was used to construct a mathematical relationship model between rainfall intensity and the historical precipitations character. First of all, we calculated the Variation Function of each site in time dimension, and analyzed the relationship between the variation function values under different time scales and the distribution of rainfall. Then we estimated the spatial distribution of precipitation by choosing variation function values under different time scales as collaborative variables. At last, we evaluated the accuracy of the Temporal Sequence co Kriging model and the Ordinary Kriging model by both the cross validation and the Error comparison of every Rain gauge station. In addition, according to the obvious seasonal trends of precipitation data, this paper studied the change of the Temporal Sequence co Kriging models accuracy before and after the temporal decomposition, and then summarized the characteristics of the models time scale. The research results showed that the precision of the precipitation distribution estimation improved obviously by using the Temporal Sequence co Kriging method than the Ordinary Kriging method both in the overall accuracy and in the single point accuracy. 〖JP2〗Data after time series decomposing could steadily improve the estimation precision than before. It was feasible to use the Temporal Sequence co Kriging method for the precipitation distribution estimation under the condition which is full of historical precipitation data and short of other precipitation information. Meanwhile, the calculation of the Variation Function in time dimension was very easy, and could be adjusted dynamically according to the rain gauge observation site real time data. Because the Temporal Sequence co Kriging method had simple, strong timeliness with the characteristics of high precision, it was an effective method for the precipitation distribution estimation based on the rain gauge observation data