通过隐马尔可夫模型(Hidden Markov Model,HMM)对安徽省降水规律及特征进行分析模拟,以验证其在区域性降水方面的适用性。采用包含4个隐式状态的HMM对省内6个主要城市的多年日降水数据序列进行拟合。用贝叶斯信息准则(Bayesian Information Criterions,BIC)确定模型中隐式状态数量,用Baum-Welch算法训练得到最优模型参数,用Viterbi算法确定模型中最优状态序列。采用上述方法模拟安徽省6个城市在1960—2009年夏季共50个时段的降水情况。前40 a用于模型分析训练,后10 a用于模型验证及评价,结果表明HMM能更好地模拟降水特征,具有较高的实用性。
Abstract
The laws and characteristics of precipitation in Anhui Province were analyzed and simulated using the Hidden Markov Model (HMM) to verify its applicability in regional precipitation. HMM with four implicit states was employed to fit the daily precipitation data sequence of many years in six major cities in the province. Bayesian Information Criterion was adopted to determine the implicit state quantity, the Baum-Welch algorithm to train and obtain the optimal model parameters, and the Viterbi algorithm to determine the optimal sequence of the model states. The above methods were adopted to simulate the precipitation in the summer of 1960-2009 in six cities of Anhui Province. The first 4-decade was for model training and analyzing, and the later 1 decade for model validation and evaluation. Results showed that HMM is of high practicability by better simulating rainfall characteristics.
关键词
隐马尔可夫模型 /
隐式状态 /
降水特征 /
贝叶斯信息准则 /
Baum-Welch算法 /
Viterbi算法
Key words
hidden Markov model /
implicit state /
precipitation characteristics /
Bayesian information criterion /
Baum-Welch algorithm /
Viterbi algorithm
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] HUGHES J P, GUTTORP P, CHARLES S P. A Non-homogeneous Hidden Markov Model for Precipitation Occurrence[J]. Applied Statistics, 1999, 48(1): 15-30.
[2] THYER M, KUCZERA G. Modeling Long-term Persistence in Hydro-climatic Time Series Using a Hidden State Markov Model[J]. Water Resources Research, 2000, 36(11): 3301-3310.
[3] THYER M, KUCZERA G. A Hidden Markov Model for Modelling Long-term Persistence in Multi-site Rainfall Time Series 1. Model Calibration Using a Bayesian Approach[J]. Journal of Hydrology, 2003, 275(1/2):12-26.
[4] HOLZMANN H, MUNK A, SUSTER M, et al. Hidden Markov Models for Circular and Linear-circular Time Series[J]. Environmental and Ecological Statistics, 2006, 13(3): 325-347.
[5] MARES C, MARES I, HUEBENER H, et al. A Hidden Markov Model Applied to the Daily Spring Precipitation over the Danube Basin[J]. Advances in Meteorology, 2014: 1-11. http://dx.doi.org/10.1155/2014/237247.
[6] 袁宏永, 张毛磊, 杨 锐,等.基于隐式Markov方法的福建省降水预测[J].清华大学学报(自然科学版),2010,50(6):887-890.
[7] BELLONE E, HUGHES J P, GUTTORP P. A Hidden Markov Model for Downscaling Synoptic Atmospheric Patterns to Precipitation Amounts[J]. Climate Research, 2000,15(1): 1-12.
[8] 陈 健,杜晓宾,方 茸,等. 安徽省夏季降水区域特征[J]. 气象科技,2012,40(5):794-798.
[9] 江俊杰,孙卫国.1959—2007年安徽省降水时空变化特征分析[J].中国农业气象,2012, 33(1):27-33.
[10]BETRO B, BODINI A, COSSU Q A. Using a Hidden Markov Model to Analyse Extreme Rainfall Events in Central-East Sardinia[J]. Environmetrics, 2008, 19(7): 702-713.
[11]KEHAGIAS A. A Hidden Markov Model Segmentation Procedure for Hydrological and Environmental Time Series[J]. Stochastic Environmental Research and Risk Assessment, 2004, 18: 117-130.
[12]RAMESH N I, ONOF C. A Class of Hidden Markov Models for Regional Average Rainfall[J]. Hydrological Sciences Journal,2014, 59(9): 23-38.
[13]SANSOM J. Hidden Markov Model for Rainfall Using Breakpoint Data[J]. Journal of Climate, 1997, 11(1): 42-53.
[14]LEROUX B G. Maximum-likelihood Estimation for Hidden Markov Models[J]. Stochastic Processes and Their Applications, 1992, 40(1): 127-143.
[15]张 茜,梁秀娟,杜 川. 基于无偏灰色马尔可夫链的吉林省降水量预测[J]. 吉林大学学报(地球科学版),2014,(6):1973-1979.
PDF(1755 KB)
