| 引用本文: | 张钦,刘赛艳,解阳阳,等.基于交叉验证法与马尔科夫链的年径流丰枯分类可靠性研究[J].灌溉排水学报,0,():-. |
| ZHANG Qin,LIU Saiyan,XIE Yangyang,et al.基于交叉验证法与马尔科夫链的年径流丰枯分类可靠性研究[J].灌溉排水学报,0,():-. |
|
| 摘要: |
| 【目的】提出一种用于检验年径流丰枯分类可靠性的方法，为划分流域年径流丰枯类别提供科学依据。【方法】从稳定性和可预测性2个方面分析年径流丰枯分类的可靠性，提出基于交叉验证法和马尔科夫链的检验方法；以常用的均值标准差法、灰色关联分析法和集对分析法为代表，采用交叉验证法和马尔科夫链分别研究各方法所得分类结果的稳定性和可预测性；建立分类差异指数和转移概率差值，用于评价不同分类方法划分年径流丰枯类别的稳定性和可预测性；最后以黄河上游唐乃亥站年径流序列（1956—2021年）为例进行验证。【结果】（1）基于交叉验证法和马尔科夫链，不同分类方法得到的年径流丰枯状态均存在差异，表明不同分类方法划分和预测年径流丰枯状态的稳定性和可预测性不同；（2）对唐乃亥站而言，分类差异指数表明GRA法的稳定性最好，SPA法其次，MSD法最差。转移概率差值表明GRA法的可预测性最好，SPA法其次，MSD法最差；（3）综合考虑各分类方法的稳定性和可预测性，对唐乃亥站而言，GRA法划分年径流丰枯状态最为可靠，SPA法其次，MSD法最差。【结论】不同年径流丰枯分类方法在同一流域存在可靠性差异，基于交叉验证法和马尔科夫链的方法能够有效检验各分类方法所得结果的可靠性。 |
| 关键词: 丰枯分类可靠性；交叉验证；马尔科夫链；灰色关联分析；集对分析 |
| DOI： |
| 分类号:P333 |
| 基金项目:国家自然科学基金（52009116）；江苏省自然科学基金(BK20200959；BK20200958)；中国博士后科学基金（2018M642338）；扬州市软科学研究课题（2022187） |
|
| Reliability of Classification of Annual Runoff Wet-dry State Based on Cross-validation and Markov Chain |
|
ZHANG Qin, LIU Saiyan, XIE Yangyang, XI Haichao
|
|
College of Hydraulic Science and Engineering, Yangzhou University
|
| Abstract: |
| 【Objective】A method for testing the reliabilities of annual runoff wet-dry classification results is proposed to provide a scientific basis for the classification of annual runoff wet-dry states in a watershed. 【Method】 The reliability of annual runoff wet-dry classification is analyzed in terms of both stability and predictability, and the test method based on the cross-validation method and Markov chain is proposed. Taking the commonly used mean-standard deviation method (MSD), gray relation analysis (GRA) and set-pair analysis (SPA) as representatives, the stability and predictability of the classification results obtained by each method are investigated separately using the cross-validation method and Markov chain. The classification difference and transfer probability difference indices are built to evaluate the stability and predictability of different classification results. Finally, the annual runoff series from 1956—2021 at the upper reaches of the Yellow River at Tangnaihai station is used as an example for verification.【Result】(1) The results of different classification methods based on the cross-validation method and Markov chain have differences, which indicates that the stability and predictability of different methods in classifying and predicting annual runoff wet-dry states are different. (2) For Tangnaihai station, the classification difference index indicates that the GRA method is the most stable, followed by the SPA method, and the MSD method is the worst. Transfer probability differences indicates that the GRA method is the most predictable, the SPA method the next most predictable and the MSD method the worst. (3) Considering the stability and predictability of the classification methods, the GRA method is the most reliable method for classifying annual runoff abundance and depletion, followed by the SPA method, and the MSD method is the worst.【Conclusion】The reliability of different annual runoff wet-dry classification methods varies in the same watershed, and the method based on the cross-validation method and Markov chain can effectively test the reliability of the results of each classification method. |
| Key words: reliability of wet-dry classification; cross validation; Markov chain; gray relation analysis; set pair analysis |
