一种马尔可夫链年降水量模型计算方法

Method for calculating annual precipitation models using a Markov chain

为解决马尔可夫链年降水量预测模型计算过程中缺乏一致性的问题,在加权马氏链年降水量模型基础上,提出降水量标准分数计算函数和预测状态标准分数转移平均值估算方法。先构建状态转移标准分平均值函数关系,由计算函数估计预测状态的标准分,以估计标准分推算预测状态降水量值。本文以石家庄、济南两市多年降水量为样本进行对比验证,预测了2020年降水量状态和估算值,分析了降水量预测值不同置信区间的取值范围。计算结果表明,马氏链年降水量预测模型对2、3状态预测效果较好,对干旱或雨涝等极值降水年份预测精度较差,两市共12个计算结果中,预测状态正确率为0.75,估算标准分确定性系数大于0.89,降水量预测值平均相对误差小于9.2%。

In order to address the issue of inconsistency in the calculation process of the Markov chain annual precipitation forecasting model,this study proposes a method based on the Weighted Markov Chain for modeling annual precipitation.It introduces a precipitation standard score calculation function and an estimation method for the average standard score transition of predictive states.First,it establishes a relationship function for the average standard score transition of states.Then,it estimates the standard score of predictive states using the calculation function and derives the precipitation values of predictive states from the estimated standard scores.This study conducted a comparative validation using multi-year precipitation data from two cities,Shijiazhuang and Jinan.It predicted precipitation states and estimated values for the year 2020 and analyzed the range of confidence intervals for different precipitation forecasting values.The computational results indicate that the Markov chain annual precipitation forecasting model performs well in predicting states with 2 or 3 states but has lower accuracy for extreme precipitation years,such as drought or heavy rainfall.Out of 12 calculations for the two cities,the accuracy of state prediction was 0.75,the determinacy coefficient for estimated standard scores was greater than 0.89,and the average relative error in precipitation forecasting values was less than 9.2%.